1 /* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
2 * vim:expandtab:shiftwidth=8:tabstop=8:
5 #ifndef LLITE_INTERNAL_H
6 #define LLITE_INTERNAL_H
8 #include <linux/lustre_acl.h>
10 #ifdef CONFIG_FS_POSIX_ACL
11 # include <linux/fs.h>
13 # include <linux/xattr_acl.h>
14 #endif /* HAVE_XATTR_ACL */
15 #ifdef HAVE_LINUX_POSIX_ACL_XATTR_H
16 # include <linux/posix_acl_xattr.h>
17 #endif /* HAVE_LINUX_POSIX_ACL_XATTR_H */
18 #endif /* CONFIG_FS_POSIX_ACL */
20 #include <lustre_debug.h>
21 #include <lustre_ver.h>
22 #include <lustre_disk.h> /* for s2sbi */
23 #include <lustre_eacl.h>
29 #define LL_IT2STR(it) ((it) ? ldlm_it2str((it)->it_op) : "0")
30 #define LUSTRE_FPRIVATE(file) ((file)->private_data)
32 #ifdef HAVE_VFS_INTENT_PATCHES
33 static inline struct lookup_intent *ll_nd2it(struct nameidata *nd)
39 struct ll_dentry_data {
42 struct obd_client_handle lld_cwd_och;
43 struct obd_client_handle lld_mnt_och;
44 #ifndef HAVE_VFS_INTENT_PATCHES
45 struct lookup_intent *lld_it;
47 unsigned int lld_sa_generation;
48 cfs_waitq_t lld_waitq;
51 #define ll_d2d(de) ((struct ll_dentry_data*)((de)->d_fsdata))
53 extern struct file_operations ll_pgcache_seq_fops;
55 #define LLI_INODE_MAGIC 0x111d0de5
56 #define LLI_INODE_DEAD 0xdeadd00d
58 /* remote client permission cache */
59 #define REMOTE_PERM_HASHSIZE 16
61 /* llite setxid/access permission for user on remote client */
62 struct ll_remote_perm {
63 struct hlist_node lrp_list;
68 int lrp_access_perm; /* MAY_READ/WRITE/EXEC, this
69 is access permission with
70 lrp_fsuid/lrp_fsgid. */
74 /* MDS has an authority for the Size-on-MDS attributes. */
75 LLIF_MDS_SIZE_LOCK = (1 << 0),
76 /* Epoch close is postponed. */
77 LLIF_EPOCH_PENDING = (1 << 1),
78 /* DONE WRITING is allowed. */
79 LLIF_DONE_WRITING = (1 << 2),
80 /* Sizeon-on-MDS attributes are changed. An attribute update needs to
82 LLIF_SOM_DIRTY = (1 << 3),
83 /* File is contented */
84 LLIF_CONTENDED = (1 << 4),
85 /* Truncate uses server lock for this file */
86 LLIF_SRVLOCK = (1 << 5)
90 struct ll_inode_info {
92 struct semaphore lli_size_sem; /* protect open and change size */
93 void *lli_size_sem_owner;
94 struct semaphore lli_write_sem;
95 char *lli_symlink_name;
98 unsigned long lli_flags;
99 cfs_time_t lli_contention_time;
101 /* this lock protects posix_acl, pending_write_llaps, mmap_cnt */
103 struct list_head lli_pending_write_llaps;
104 struct list_head lli_close_list;
105 /* handle is to be sent to MDS later on done_writing and setattr.
106 * Open handle data are needed for the recovery to reconstruct
107 * the inode state on the MDS. XXX: recovery is not ready yet. */
108 struct obd_client_handle *lli_pending_och;
109 atomic_t lli_mmap_cnt;
111 /* for writepage() only to communicate to fsync */
114 struct posix_acl *lli_posix_acl;
116 /* remote permission hash */
117 struct hlist_head *lli_remote_perms;
118 unsigned long lli_rmtperm_utime;
119 struct semaphore lli_rmtperm_sem;
121 struct list_head lli_dead_list;
123 struct semaphore lli_och_sem; /* Protects access to och pointers
124 and their usage counters */
125 /* We need all three because every inode may be opened in different
127 struct obd_client_handle *lli_mds_read_och;
128 __u64 lli_open_fd_read_count;
129 struct obd_client_handle *lli_mds_write_och;
130 __u64 lli_open_fd_write_count;
131 struct obd_client_handle *lli_mds_exec_och;
132 __u64 lli_open_fd_exec_count;
134 struct inode lli_vfs_inode;
136 /* identifying fields for both metadata and data stacks. */
137 struct lu_fid lli_fid;
138 struct lov_stripe_md *lli_smd;
141 /* open count currently used by capability only, indicate whether
142 * capability needs renewal */
143 atomic_t lli_open_count;
144 struct obd_capa *lli_mds_capa;
145 struct list_head lli_oss_capas;
147 /* metadata stat-ahead */
149 * "opendir_pid" is the token when lookup/revalid -- I am the owner of
152 pid_t lli_opendir_pid;
154 * since parent-child threads can share the same @file struct,
155 * "opendir_key" is the token when dir close for case of parent exit
156 * before child -- it is me should cleanup the dir readahead. */
157 void *lli_opendir_key;
158 struct ll_statahead_info *lli_sai;
162 * Locking to guarantee consistency of non-atomic updates to long long i_size,
163 * consistency between file size and KMS, and consistency within
164 * ->lli_smd->lsm_oinfo[]'s.
166 * Implemented by ->lli_size_sem and ->lsm_sem, nested in that order.
169 void ll_inode_size_lock(struct inode *inode, int lock_lsm);
170 void ll_inode_size_unlock(struct inode *inode, int unlock_lsm);
172 // FIXME: replace the name of this with LL_I to conform to kernel stuff
173 // static inline struct ll_inode_info *LL_I(struct inode *inode)
174 static inline struct ll_inode_info *ll_i2info(struct inode *inode)
176 return container_of(inode, struct ll_inode_info, lli_vfs_inode);
179 /* default to about 40meg of readahead on a given system. That much tied
180 * up in 512k readahead requests serviced at 40ms each is about 1GB/s. */
181 #define SBI_DEFAULT_READAHEAD_MAX (40UL << (20 - CFS_PAGE_SHIFT))
183 /* default to read-ahead full files smaller than 2MB on the second read */
184 #define SBI_DEFAULT_READAHEAD_WHOLE_MAX (2UL << (20 - CFS_PAGE_SHIFT))
189 RA_STAT_DISTANT_READPAGE,
190 RA_STAT_MISS_IN_WINDOW,
191 RA_STAT_FAILED_GRAB_PAGE,
192 RA_STAT_FAILED_MATCH,
197 RA_STAT_MAX_IN_FLIGHT,
198 RA_STAT_WRONG_GRAB_PAGE,
203 unsigned long ra_cur_pages;
204 unsigned long ra_max_pages;
205 unsigned long ra_max_read_ahead_whole_pages;
206 unsigned long ra_stats[_NR_RA_STAT];
209 /* LL_HIST_MAX=32 causes an overflow */
210 #define LL_HIST_MAX 28
211 #define LL_HIST_START 12 /* buckets start at 2^12 = 4k */
212 #define LL_PROCESS_HIST_MAX 10
213 struct per_process_info {
215 struct obd_histogram pp_r_hist;
216 struct obd_histogram pp_w_hist;
219 /* pp_extents[LL_PROCESS_HIST_MAX] will hold the combined process info */
220 struct ll_rw_extents_info {
221 struct per_process_info pp_extents[LL_PROCESS_HIST_MAX + 1];
224 #define LL_OFFSET_HIST_MAX 100
225 struct ll_rw_process_info {
228 loff_t rw_range_start;
230 loff_t rw_last_file_pos;
232 size_t rw_smallest_extent;
233 size_t rw_largest_extent;
234 struct file *rw_last_file;
237 enum stats_track_type {
238 STATS_TRACK_ALL = 0, /* track all processes */
239 STATS_TRACK_PID, /* track process with this pid */
240 STATS_TRACK_PPID, /* track processes with this ppid */
241 STATS_TRACK_GID, /* track processes with this gid */
245 /* flags for sbi->ll_flags */
246 #define LL_SBI_NOLCK 0x01 /* DLM locking disabled (directio-only) */
247 #define LL_SBI_CHECKSUM 0x02 /* checksum each page as it's written */
248 #define LL_SBI_FLOCK 0x04
249 #define LL_SBI_USER_XATTR 0x08 /* support user xattr */
250 #define LL_SBI_ACL 0x10 /* support ACL */
251 #define LL_SBI_JOIN 0x20 /* support JOIN */
252 #define LL_SBI_RMT_CLIENT 0x40 /* remote client */
253 #define LL_SBI_MDS_CAPA 0x80 /* support mds capa */
254 #define LL_SBI_OSS_CAPA 0x100 /* support oss capa */
255 #define LL_SBI_LOCALFLOCK 0x200 /* Local flocks support by kernel */
256 #define LL_SBI_LRU_RESIZE 0x400 /* lru resize support */
258 /* default value for ll_sb_info->contention_time */
259 #define SBI_DEFAULT_CONTENTION_SECONDS 60
260 /* default value for lockless_truncate_enable */
261 #define SBI_DEFAULT_LOCKLESS_TRUNCATE_ENABLE 1
262 #define RCE_HASHES 32
264 struct rmtacl_ctl_entry {
265 struct list_head rce_list;
266 pid_t rce_key; /* hash key */
267 int rce_ops; /* acl operation type */
270 struct rmtacl_ctl_table {
272 struct list_head rct_entries[RCE_HASHES];
278 struct list_head ee_list;
279 pid_t ee_key; /* hash key */
280 struct lu_fid ee_fid;
281 int ee_type; /* ACL type for ACCESS or DEFAULT */
282 ext_acl_xattr_header *ee_acl;
287 struct list_head et_entries[EE_HASHES];
291 struct list_head ll_list;
292 /* this protects pglist and ra_info. It isn't safe to
293 * grab from interrupt contexts */
295 spinlock_t ll_pp_extent_lock; /* Lock for pp_extent entries */
296 spinlock_t ll_process_lock; /* Lock for ll_rw_process_info */
297 struct obd_uuid ll_sb_uuid;
298 struct obd_export *ll_md_exp;
299 struct obd_export *ll_dt_exp;
300 struct proc_dir_entry* ll_proc_root;
301 struct lu_fid ll_root_fid; /* root object fid */
304 struct list_head ll_conn_chain; /* per-conn chain of SBs */
305 struct lustre_client_ocd ll_lco;
307 struct list_head ll_orphan_dentry_list; /*please don't ask -p*/
308 struct ll_close_queue *ll_lcq;
310 struct lprocfs_stats *ll_stats; /* lprocfs stats counter */
312 unsigned long ll_async_page_max;
313 unsigned long ll_async_page_count;
314 unsigned long ll_pglist_gen;
315 struct list_head ll_pglist; /* all pages (llap_pglist_item) */
317 unsigned ll_contention_time; /* seconds */
318 unsigned ll_lockless_truncate_enable; /* true/false */
320 struct ll_ra_info ll_ra_info;
321 unsigned int ll_namelen;
322 struct file_operations *ll_fop;
324 #ifdef HAVE_EXPORT___IGET
325 struct list_head ll_deathrow; /* inodes to be destroyed (b1443) */
326 spinlock_t ll_deathrow_lock;
328 /* =0 - hold lock over whole read/write
329 * >0 - max. chunk to be read/written w/o lock re-acquiring */
330 unsigned long ll_max_rw_chunk;
333 struct ll_rw_extents_info ll_rw_extents_info;
334 int ll_extent_process_count;
335 struct ll_rw_process_info ll_rw_process_info[LL_PROCESS_HIST_MAX];
336 unsigned int ll_offset_process_count;
337 struct ll_rw_process_info ll_rw_offset_info[LL_OFFSET_HIST_MAX];
338 unsigned int ll_rw_offset_entry_count;
339 enum stats_track_type ll_stats_track_type;
340 int ll_stats_track_id;
343 /* metadata stat-ahead */
344 unsigned int ll_sa_max; /* max statahead RPCs */
345 unsigned int ll_sa_wrong; /* statahead thread stopped for
347 unsigned int ll_sa_total; /* statahead thread started
349 unsigned long long ll_sa_blocked; /* ls count waiting for
351 unsigned long long ll_sa_cached; /* ls count got in cache */
352 unsigned long long ll_sa_hit; /* hit count */
353 unsigned long long ll_sa_miss; /* miss count */
355 dev_t ll_sdev_orig; /* save s_dev before assign for
357 struct rmtacl_ctl_table ll_rct;
358 struct eacl_table ll_et;
361 #define LL_DEFAULT_MAX_RW_CHUNK (32 * 1024 * 1024)
366 struct task_struct *lrr_reader;
367 struct list_head lrr_linkage;
371 * per file-descriptor read-ahead data.
373 struct ll_readahead_state {
376 * index of the last page that read(2) needed and that wasn't in the
377 * cache. Used by ras_update() to detect seeks.
379 * XXX nikita: if access seeks into cached region, Lustre doesn't see
382 unsigned long ras_last_readpage;
384 * number of pages read after last read-ahead window reset. As window
385 * is reset on each seek, this is effectively a number of consecutive
386 * accesses. Maybe ->ras_accessed_in_window is better name.
388 * XXX nikita: window is also reset (by ras_update()) when Lustre
389 * believes that memory pressure evicts read-ahead pages. In that
390 * case, it probably doesn't make sense to expand window to
391 * PTLRPC_MAX_BRW_PAGES on the third access.
393 unsigned long ras_consecutive_pages;
395 * number of read requests after the last read-ahead window reset
396 * As window is reset on each seek, this is effectively the number
397 * on consecutive read request and is used to trigger read-ahead.
399 unsigned long ras_consecutive_requests;
401 * Parameters of current read-ahead window. Handled by
402 * ras_update(). On the initial access to the file or after a seek,
403 * window is reset to 0. After 3 consecutive accesses, window is
404 * expanded to PTLRPC_MAX_BRW_PAGES. Afterwards, window is enlarged by
405 * PTLRPC_MAX_BRW_PAGES chunks up to ->ra_max_pages.
407 unsigned long ras_window_start, ras_window_len;
409 * Where next read-ahead should start at. This lies within read-ahead
410 * window. Read-ahead window is read in pieces rather than at once
411 * because: 1. lustre limits total number of pages under read-ahead by
412 * ->ra_max_pages (see ll_ra_count_get()), 2. client cannot read pages
413 * not covered by DLM lock.
415 unsigned long ras_next_readahead;
417 * Total number of ll_file_read requests issued, reads originating
418 * due to mmap are not counted in this total. This value is used to
419 * trigger full file read-ahead after multiple reads to a small file.
421 unsigned long ras_requests;
423 * Page index with respect to the current request, these value
424 * will not be accurate when dealing with reads issued via mmap.
426 unsigned long ras_request_index;
428 * list of struct ll_ra_read's one per read(2) call current in
429 * progress against this file descriptor. Used by read-ahead code,
430 * protected by ->ras_lock.
432 struct list_head ras_read_beads;
434 * The following 3 items are used for detecting the stride I/O
436 * In stride I/O mode,
437 * ...............|-----data-----|****gap*****|--------|******|....
438 * offset |-stride_pages-|-stride_gap-|
439 * ras_stride_offset = offset;
440 * ras_stride_length = stride_pages + stride_gap;
441 * ras_stride_pages = stride_pages;
442 * Note: all these three items are counted by pages.
444 unsigned long ras_stride_length;
445 unsigned long ras_stride_pages;
446 pgoff_t ras_stride_offset;
448 * number of consecutive stride request count, and it is similar as
449 * ras_consecutive_requests, but used for stride I/O mode.
450 * Note: only more than 2 consecutive stride request are detected,
451 * stride read-ahead will be enable
453 unsigned long ras_consecutive_stride_requests;
459 extern cfs_mem_cache_t *ll_file_data_slab;
460 struct lustre_handle;
461 struct ll_file_data {
462 struct ll_readahead_state fd_ras;
464 struct lustre_handle fd_cwlockh;
465 unsigned long fd_gid;
466 struct ll_file_dir fd_dir;
470 struct lov_stripe_md;
472 extern spinlock_t inode_lock;
474 extern struct proc_dir_entry *proc_lustre_fs_root;
476 static inline struct inode *ll_info2i(struct ll_inode_info *lli)
478 return &lli->lli_vfs_inode;
482 struct inode *icbd_parent;
483 struct dentry **icbd_childp;
487 __u32 ll_i2suppgid(struct inode *i);
488 void ll_i2gids(__u32 *suppgids, struct inode *i1,struct inode *i2);
490 #define LLAP_MAGIC 98764321
492 extern cfs_mem_cache_t *ll_async_page_slab;
493 extern size_t ll_async_page_slab_size;
494 struct ll_async_page {
496 /* only trust these if the page lock is providing exclusion */
497 unsigned int llap_write_queued:1,
498 llap_defer_uptodate:1,
503 llap_lockless_io_page:1;
505 struct page *llap_page;
506 struct list_head llap_pending_write;
507 struct list_head llap_pglist_item;
508 /* checksum for paranoid I/O debugging */
513 * enumeration of llap_from_page() call-sites. Used to export statistics in
514 * /proc/fs/lustre/llite/fsN/dump_page_cache.
517 LLAP_ORIGIN_UNKNOWN = 0,
518 LLAP_ORIGIN_READPAGE,
519 LLAP_ORIGIN_READAHEAD,
520 LLAP_ORIGIN_COMMIT_WRITE,
521 LLAP_ORIGIN_WRITEPAGE,
522 LLAP_ORIGIN_REMOVEPAGE,
523 LLAP_ORIGIN_LOCKLESS_IO,
526 extern char *llap_origins[];
528 #ifdef HAVE_REGISTER_CACHE
529 #define ll_register_cache(cache) register_cache(cache)
530 #define ll_unregister_cache(cache) unregister_cache(cache)
532 #define ll_register_cache(cache) do {} while (0)
533 #define ll_unregister_cache(cache) do {} while (0)
536 void ll_ra_read_in(struct file *f, struct ll_ra_read *rar);
537 void ll_ra_read_ex(struct file *f, struct ll_ra_read *rar);
538 struct ll_ra_read *ll_ra_read_get(struct file *f);
540 /* llite/lproc_llite.c */
542 int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
543 struct super_block *sb, char *osc, char *mdc);
544 void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi);
545 void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count);
546 void lprocfs_llite_init_vars(struct lprocfs_static_vars *lvars);
548 static inline int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
549 struct super_block *sb, char *osc, char *mdc){return 0;}
550 static inline void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi) {}
551 static void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count) {}
552 static void lprocfs_llite_init_vars(struct lprocfs_static_vars *lvars)
554 memset(lvars, 0, sizeof(*lvars));
560 static inline void ll_put_page(struct page *page)
563 page_cache_release(page);
566 extern struct file_operations ll_dir_operations;
567 extern struct inode_operations ll_dir_inode_operations;
568 struct page *ll_get_dir_page(struct inode *dir, __u64 hash, int exact,
569 struct ll_dir_chain *chain);
571 int ll_objects_destroy(struct ptlrpc_request *request,
573 struct inode *ll_iget(struct super_block *sb, ino_t hash,
574 struct lustre_md *lic);
575 int ll_md_blocking_ast(struct ldlm_lock *, struct ldlm_lock_desc *,
576 void *data, int flag);
577 #ifndef HAVE_VFS_INTENT_PATCHES
578 struct lookup_intent *ll_convert_intent(struct open_intent *oit,
581 int ll_lookup_it_finish(struct ptlrpc_request *request,
582 struct lookup_intent *it, void *data);
583 void ll_lookup_finish_locks(struct lookup_intent *it, struct dentry *dentry);
586 int ll_prepare_write(struct file *, struct page *, unsigned from, unsigned to);
587 int ll_commit_write(struct file *, struct page *, unsigned from, unsigned to);
588 int ll_writepage(struct page *page);
589 void ll_inode_fill_obdo(struct inode *inode, int cmd, struct obdo *oa);
590 int ll_ap_completion(void *data, int cmd, struct obdo *oa, int rc);
591 int llap_shrink_cache(struct ll_sb_info *sbi, int shrink_fraction);
592 struct ll_async_page *llap_from_page(struct page *page, unsigned origin);
593 extern struct cache_definition ll_cache_definition;
594 void ll_removepage(struct page *page);
595 int ll_readpage(struct file *file, struct page *page);
596 struct ll_async_page *llap_cast_private(struct page *page);
597 void ll_readahead_init(struct inode *inode, struct ll_readahead_state *ras);
598 void ll_ra_accounting(struct ll_async_page *llap,struct address_space *mapping);
599 void ll_truncate(struct inode *inode);
600 int ll_file_punch(struct inode *, loff_t, int);
601 ssize_t ll_file_lockless_io(struct file *, char *, size_t, loff_t *, int);
602 void ll_clear_file_contended(struct inode*);
603 int ll_sync_page_range(struct inode *, struct address_space *, loff_t, size_t);
606 extern struct file_operations ll_file_operations;
607 extern struct file_operations ll_file_operations_flock;
608 extern struct file_operations ll_file_operations_noflock;
609 extern struct inode_operations ll_file_inode_operations;
610 extern int ll_inode_revalidate_it(struct dentry *, struct lookup_intent *);
611 extern int ll_have_md_lock(struct inode *inode, __u64 bits);
612 extern ldlm_mode_t ll_take_md_lock(struct inode *inode, __u64 bits,
613 struct lustre_handle *lockh);
614 int ll_region_mapped(unsigned long addr, size_t count);
615 int ll_extent_lock(struct ll_file_data *, struct inode *,
616 struct lov_stripe_md *, int mode, ldlm_policy_data_t *,
617 struct lustre_handle *, int ast_flags);
618 int ll_extent_unlock(struct ll_file_data *, struct inode *,
619 struct lov_stripe_md *, int mode, struct lustre_handle *);
620 int ll_file_open(struct inode *inode, struct file *file);
621 int ll_file_release(struct inode *inode, struct file *file);
622 int ll_lsm_getattr(struct obd_export *, struct lov_stripe_md *, struct obdo *);
623 int ll_local_size(struct inode *inode);
624 int ll_glimpse_ioctl(struct ll_sb_info *sbi,
625 struct lov_stripe_md *lsm, lstat_t *st);
626 int ll_glimpse_size(struct inode *inode, int ast_flags);
627 int ll_local_open(struct file *file,
628 struct lookup_intent *it, struct ll_file_data *fd,
629 struct obd_client_handle *och);
630 int ll_release_openhandle(struct dentry *, struct lookup_intent *);
631 int ll_md_close(struct obd_export *md_exp, struct inode *inode,
633 int ll_md_real_close(struct inode *inode, int flags);
634 void ll_epoch_close(struct inode *inode, struct md_op_data *op_data,
635 struct obd_client_handle **och, unsigned long flags);
636 int ll_sizeonmds_update(struct inode *inode, struct md_open_data *data,
637 struct lustre_handle *fh, __u64 ioepoch);
638 int ll_inode_getattr(struct inode *inode, struct obdo *obdo);
639 int ll_md_setattr(struct inode *inode, struct md_op_data *op_data,
640 struct md_open_data **mod);
641 void ll_pack_inode2opdata(struct inode *inode, struct md_op_data *op_data,
642 struct lustre_handle *fh);
643 extern void ll_rw_stats_tally(struct ll_sb_info *sbi, pid_t pid, struct file
644 *file, size_t count, int rw);
645 int ll_getattr_it(struct vfsmount *mnt, struct dentry *de,
646 struct lookup_intent *it, struct kstat *stat);
647 int ll_getattr(struct vfsmount *mnt, struct dentry *de, struct kstat *stat);
648 struct ll_file_data *ll_file_data_get(void);
649 int ll_inode_permission(struct inode *inode, int mask, struct nameidata *nd);
650 int ll_lov_setstripe_ea_info(struct inode *inode, struct file *file,
651 int flags, struct lov_user_md *lum,
653 int ll_lov_getstripe_ea_info(struct inode *inode, const char *filename,
654 struct lov_mds_md **lmm, int *lmm_size,
655 struct ptlrpc_request **request);
656 int ll_dir_setstripe(struct inode *inode, struct lov_user_md *lump,
658 int ll_dir_getstripe(struct inode *inode, struct lov_mds_md **lmm,
659 int *lmm_size, struct ptlrpc_request **request);
660 void ll_pin_extent_cb(void *data);
661 int ll_page_removal_cb(void *data, int discard);
662 int ll_extent_lock_cancel_cb(struct ldlm_lock *lock, struct ldlm_lock_desc *new,
663 void *data, int flag);
666 extern struct dentry_operations ll_init_d_ops;
667 extern struct dentry_operations ll_d_ops;
668 extern struct dentry_operations ll_fini_d_ops;
669 void ll_intent_drop_lock(struct lookup_intent *);
670 void ll_intent_release(struct lookup_intent *);
671 int ll_drop_dentry(struct dentry *dentry);
672 extern void ll_set_dd(struct dentry *de);
673 int ll_drop_dentry(struct dentry *dentry);
674 void ll_unhash_aliases(struct inode *);
675 void ll_frob_intent(struct lookup_intent **itp, struct lookup_intent *deft);
676 void ll_lookup_finish_locks(struct lookup_intent *it, struct dentry *dentry);
677 int ll_dcompare(struct dentry *parent, struct qstr *d_name, struct qstr *name);
678 int ll_revalidate_it_finish(struct ptlrpc_request *request,
679 struct lookup_intent *it, struct dentry *de);
681 /* llite/llite_lib.c */
682 extern struct super_operations lustre_super_operations;
684 char *ll_read_opt(const char *opt, char *data);
685 void ll_lli_init(struct ll_inode_info *lli);
686 int ll_fill_super(struct super_block *sb);
687 void ll_put_super(struct super_block *sb);
688 void ll_kill_super(struct super_block *sb);
689 struct inode *ll_inode_from_lock(struct ldlm_lock *lock);
690 void ll_clear_inode(struct inode *inode);
691 int ll_setattr_raw(struct inode *inode, struct iattr *attr);
692 int ll_setattr(struct dentry *de, struct iattr *attr);
693 #ifndef HAVE_STATFS_DENTRY_PARAM
694 int ll_statfs(struct super_block *sb, struct kstatfs *sfs);
696 int ll_statfs(struct dentry *de, struct kstatfs *sfs);
698 int ll_statfs_internal(struct super_block *sb, struct obd_statfs *osfs,
699 __u64 max_age, __u32 flags);
700 void ll_update_inode(struct inode *inode, struct lustre_md *md);
701 void ll_read_inode2(struct inode *inode, void *opaque);
702 void ll_delete_inode(struct inode *inode);
703 int ll_iocontrol(struct inode *inode, struct file *file,
704 unsigned int cmd, unsigned long arg);
705 int ll_flush_ctx(struct inode *inode);
706 #ifdef HAVE_UMOUNTBEGIN_VFSMOUNT
707 void ll_umount_begin(struct vfsmount *vfsmnt, int flags);
709 void ll_umount_begin(struct super_block *sb);
711 int ll_remount_fs(struct super_block *sb, int *flags, char *data);
712 int ll_prep_inode(struct inode **inode, struct ptlrpc_request *req,
713 struct super_block *);
714 void lustre_dump_dentry(struct dentry *, int recur);
715 void lustre_dump_inode(struct inode *);
716 struct ll_async_page *llite_pglist_next_llap(struct ll_sb_info *sbi,
717 struct list_head *list);
718 int ll_obd_statfs(struct inode *inode, void *arg);
719 int ll_get_max_mdsize(struct ll_sb_info *sbi, int *max_mdsize);
720 int ll_process_config(struct lustre_cfg *lcfg);
721 struct md_op_data *ll_prep_md_op_data(struct md_op_data *op_data,
722 struct inode *i1, struct inode *i2,
723 const char *name, int namelen,
724 int mode, __u32 opc, void *data);
725 void ll_finish_md_op_data(struct md_op_data *op_data);
727 /* llite/llite_nfs.c */
728 extern struct export_operations lustre_export_operations;
729 __u32 get_uuid2int(const char *name, int len);
730 struct dentry *ll_fh_to_dentry(struct super_block *sb, __u32 *data, int len,
731 int fhtype, int parent);
732 int ll_dentry_to_fh(struct dentry *, __u32 *datap, int *lenp, int need_parent);
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 {
749 struct list_head lcq_head;
750 wait_queue_head_t lcq_waitq;
751 struct completion lcq_comp;
755 void llap_write_pending(struct inode *inode, struct ll_async_page *llap);
756 int llap_write_complete(struct inode *inode, struct ll_async_page *llap);
757 int ll_inode_dirty(struct inode *inode, unsigned long flags);
758 void ll_queue_done_writing(struct inode *inode, unsigned long flags);
759 void ll_close_thread_shutdown(struct ll_close_queue *lcq);
760 int ll_close_thread_start(struct ll_close_queue **lcq_ret);
762 /* llite/llite_mmap.c */
763 typedef struct rb_root rb_root_t;
764 typedef struct rb_node rb_node_t;
766 struct ll_lock_tree_node;
767 struct ll_lock_tree {
769 struct list_head lt_locked_list;
770 struct ll_file_data *lt_fd;
773 int ll_teardown_mmaps(struct address_space *mapping, __u64 first, __u64 last);
774 int ll_file_mmap(struct file * file, struct vm_area_struct * vma);
775 struct ll_lock_tree_node * ll_node_from_inode(struct inode *inode, __u64 start,
776 __u64 end, ldlm_mode_t mode);
777 int ll_tree_lock(struct ll_lock_tree *tree,
778 struct ll_lock_tree_node *first_node,
779 const char *buf, size_t count, int ast_flags);
780 int ll_tree_unlock(struct ll_lock_tree *tree);
782 #define ll_s2sbi(sb) (s2lsi(sb)->lsi_llsbi)
784 static inline __u64 ll_ts2u64(struct timespec *time)
786 __u64 t = time->tv_sec;
790 /* don't need an addref as the sb_info should be holding one */
791 static inline struct obd_export *ll_s2dtexp(struct super_block *sb)
793 return ll_s2sbi(sb)->ll_dt_exp;
796 /* don't need an addref as the sb_info should be holding one */
797 static inline struct obd_export *ll_s2mdexp(struct super_block *sb)
799 return ll_s2sbi(sb)->ll_md_exp;
802 static inline struct client_obd *sbi2mdc(struct ll_sb_info *sbi)
804 struct obd_device *obd = sbi->ll_md_exp->exp_obd;
810 // FIXME: replace the name of this with LL_SB to conform to kernel stuff
811 static inline struct ll_sb_info *ll_i2sbi(struct inode *inode)
813 return ll_s2sbi(inode->i_sb);
816 static inline struct obd_export *ll_i2dtexp(struct inode *inode)
818 return ll_s2dtexp(inode->i_sb);
821 static inline struct obd_export *ll_i2mdexp(struct inode *inode)
823 return ll_s2mdexp(inode->i_sb);
826 static inline struct lu_fid *ll_inode2fid(struct inode *inode)
829 LASSERT(inode != NULL);
830 fid = &ll_i2info(inode)->lli_fid;
831 LASSERT(fid_is_igif(fid) || fid_ver(fid) == 0);
835 static inline int ll_mds_max_easize(struct super_block *sb)
837 return sbi2mdc(ll_s2sbi(sb))->cl_max_mds_easize;
840 static inline __u64 ll_file_maxbytes(struct inode *inode)
842 return ll_i2info(inode)->lli_maxbytes;
846 int ll_setxattr(struct dentry *dentry, const char *name,
847 const void *value, size_t size, int flags);
848 ssize_t ll_getxattr(struct dentry *dentry, const char *name,
849 void *buffer, size_t size);
850 ssize_t ll_listxattr(struct dentry *dentry, char *buffer, size_t size);
851 int ll_removexattr(struct dentry *dentry, const char *name);
853 /* llite/remote_perm.c */
854 extern cfs_mem_cache_t *ll_remote_perm_cachep;
855 extern cfs_mem_cache_t *ll_rmtperm_hash_cachep;
857 struct hlist_head *alloc_rmtperm_hash(void);
858 void free_rmtperm_hash(struct hlist_head *hash);
859 int ll_update_remote_perm(struct inode *inode, struct mdt_remote_perm *perm);
860 int lustre_check_remote_perm(struct inode *inode, int mask);
862 /* llite/llite_fid.c */
863 ino_t ll_fid_build_ino(struct ll_sb_info *sbi, struct lu_fid *fid);
865 /* llite/llite_capa.c */
866 extern cfs_timer_t ll_capa_timer;
868 int ll_capa_thread_start(void);
869 void ll_capa_thread_stop(void);
870 void ll_capa_timer_callback(unsigned long unused);
872 struct obd_capa *ll_add_capa(struct inode *inode, struct obd_capa *ocapa);
873 int ll_update_capa(struct obd_capa *ocapa, struct lustre_capa *capa);
875 void ll_capa_open(struct inode *inode);
876 void ll_capa_close(struct inode *inode);
878 struct obd_capa *ll_mdscapa_get(struct inode *inode);
879 struct obd_capa *ll_osscapa_get(struct inode *inode, __u64 opc);
881 void ll_truncate_free_capa(struct obd_capa *ocapa);
882 void ll_clear_inode_capas(struct inode *inode);
883 void ll_print_capa_stat(struct ll_sb_info *sbi);
885 /* llite/llite_rmtacl.c */
886 #ifdef CONFIG_FS_POSIX_ACL
887 obd_valid rce_ops2valid(int ops);
888 struct rmtacl_ctl_entry *rct_search(struct rmtacl_ctl_table *rct, pid_t key);
889 int rct_add(struct rmtacl_ctl_table *rct, pid_t key, int ops);
890 int rct_del(struct rmtacl_ctl_table *rct, pid_t key);
891 void rct_init(struct rmtacl_ctl_table *rct);
892 void rct_fini(struct rmtacl_ctl_table *rct);
894 void ee_free(struct eacl_entry *ee);
895 int ee_add(struct eacl_table *et, pid_t key, struct lu_fid *fid, int type,
896 ext_acl_xattr_header *header);
897 struct eacl_entry *et_search_del(struct eacl_table *et, pid_t key,
898 struct lu_fid *fid, int type);
899 void et_search_free(struct eacl_table *et, pid_t key);
900 void et_init(struct eacl_table *et);
901 void et_fini(struct eacl_table *et);
906 #define LL_SA_RPC_MIN 2
907 #define LL_SA_RPC_DEF 32
908 #define LL_SA_RPC_MAX 8192
910 /* per inode struct, for dir only */
911 struct ll_statahead_info {
912 struct inode *sai_inode;
913 unsigned int sai_generation; /* generation for statahead */
914 atomic_t sai_refcount; /* when access this struct, hold
916 unsigned int sai_sent; /* stat requests sent count */
917 unsigned int sai_replied; /* stat requests which received
919 unsigned int sai_max; /* max ahead of lookup */
920 unsigned int sai_index; /* index of statahead entry */
921 unsigned int sai_hit; /* hit count */
922 unsigned int sai_miss; /* miss count:
923 * for "ls -al" case, it includes
924 * hidden dentry miss;
925 * for "ls -l" case, it does not
926 * include hidden dentry miss.
927 * "sai_miss_hidden" is used for
930 unsigned int sai_consecutive_miss; /* consecutive miss */
931 unsigned int sai_miss_hidden;/* "ls -al", but first dentry
932 * is not a hidden one */
933 unsigned int sai_skip_hidden;/* skipped hidden dentry count */
934 unsigned int sai_ls_all:1; /* "ls -al", do stat-ahead for
936 cfs_waitq_t sai_waitq; /* stat-ahead wait queue */
937 struct ptlrpc_thread sai_thread; /* stat-ahead thread */
938 struct list_head sai_entries; /* stat-ahead entries */
941 int do_statahead_enter(struct inode *dir, struct dentry **dentry, int lookup);
942 void ll_statahead_exit(struct dentry *dentry, int result);
943 void ll_stop_statahead(struct inode *inode, void *key);
946 void ll_d_wakeup(struct dentry *dentry)
948 struct ll_dentry_data *lld = ll_d2d(dentry);
950 LASSERT(dentry->d_op != &ll_init_d_ops);
952 cfs_waitq_broadcast(&lld->lld_waitq);
956 int ll_statahead_enter(struct inode *dir, struct dentry **dentryp, int lookup)
958 struct ll_sb_info *sbi = ll_i2sbi(dir);
959 struct ll_inode_info *lli = ll_i2info(dir);
960 struct ll_dentry_data *ldd = ll_d2d(*dentryp);
962 if (sbi->ll_sa_max == 0)
965 /* not the same process, don't statahead */
966 if (lli->lli_opendir_pid != cfs_curproc_pid())
970 * When "ls" a dentry, the system trigger more than once "revalidate" or
971 * "lookup", for "getattr", for "getxattr", and maybe for others.
972 * Under patchless client mode, the operation intent is not accurate,
973 * it maybe misguide the statahead thread. For example:
974 * The "revalidate" call for "getattr" and "getxattr" of a dentry maybe
975 * have the same operation intent -- "IT_GETATTR".
976 * In fact, one dentry should has only one chance to interact with the
977 * statahead thread, otherwise the statahead windows will be confused.
978 * The solution is as following:
979 * Assign "lld_sa_generation" with "sai_generation" when a dentry
980 * "IT_GETATTR" for the first time, and the subsequent "IT_GETATTR"
981 * will bypass interacting with statahead thread for checking:
982 * "lld_sa_generation == lli_sai->sai_generation"
984 if (ldd && lli->lli_sai &&
985 ldd->lld_sa_generation == lli->lli_sai->sai_generation)
988 return do_statahead_enter(dir, dentryp, lookup);
991 /* llite ioctl register support rountine */
998 #define LLIOC_MAX_CMD 256
1001 * Rules to write a callback function:
1004 * @magic: Dynamic ioctl call routine will feed this vaule with the pointer
1005 * returned to ll_iocontrol_register. Callback functions should use this
1006 * data to check the potential collasion of ioctl cmd. If collasion is
1007 * found, callback function should return LLIOC_CONT.
1008 * @rcp: The result of ioctl command.
1011 * If @magic matches the pointer returned by ll_iocontrol_data, the
1012 * callback should return LLIOC_STOP; return LLIOC_STOP otherwise.
1014 typedef enum llioc_iter (*llioc_callback_t)(struct inode *inode,
1015 struct file *file, unsigned int cmd, unsigned long arg,
1016 void *magic, int *rcp);
1018 enum llioc_iter ll_iocontrol_call(struct inode *inode, struct file *file,
1019 unsigned int cmd, unsigned long arg, int *rcp);
1021 /* export functions */
1022 /* Register ioctl block dynamatically for a regular file.
1024 * @cmd: the array of ioctl command set
1025 * @count: number of commands in the @cmd
1026 * @cb: callback function, it will be called if an ioctl command is found to
1027 * belong to the command list @cmd.
1030 * A magic pointer will be returned if success;
1031 * otherwise, NULL will be returned.
1033 void *ll_iocontrol_register(llioc_callback_t cb, int count, unsigned int *cmd);
1034 void ll_iocontrol_unregister(void *magic);
1038 #endif /* LLITE_INTERNAL_H */