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 (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved.
30 * Use is subject to license terms.
33 * This file is part of Lustre, http://www.lustre.org/
34 * Lustre is a trademark of Sun Microsystems, Inc.
37 #ifndef LLITE_INTERNAL_H
38 #define LLITE_INTERNAL_H
40 #include <lustre_acl.h>
42 #include <lustre_debug.h>
43 #include <lustre_ver.h>
44 #include <lustre_disk.h> /* for s2sbi */
45 #include <lustre_eacl.h>
47 /* for struct cl_lock_descr and struct cl_io */
48 #include <cl_object.h>
55 #ifndef DCACHE_LUSTRE_INVALID
56 #define DCACHE_LUSTRE_INVALID 0x4000000
59 #define LL_IT2STR(it) ((it) ? ldlm_it2str((it)->it_op) : "0")
60 #define LUSTRE_FPRIVATE(file) ((file)->private_data)
62 struct ll_dentry_data {
65 struct obd_client_handle lld_cwd_och;
66 struct obd_client_handle lld_mnt_och;
67 struct lookup_intent *lld_it;
68 unsigned int lld_sa_generation;
71 #define ll_d2d(de) ((struct ll_dentry_data*)((de)->d_fsdata))
73 extern struct file_operations ll_pgcache_seq_fops;
75 #define LLI_INODE_MAGIC 0x111d0de5
76 #define LLI_INODE_DEAD 0xdeadd00d
78 /* remote client permission cache */
79 #define REMOTE_PERM_HASHSIZE 16
81 struct ll_getname_data {
82 char *lgd_name; /* points to a buffer with NAME_MAX+1 size */
83 struct lu_fid lgd_fid; /* target fid we are looking for */
84 int lgd_found; /* inode matched? */
87 /* llite setxid/access permission for user on remote client */
88 struct ll_remote_perm {
89 cfs_hlist_node_t lrp_list;
94 int lrp_access_perm; /* MAY_READ/WRITE/EXEC, this
95 is access permission with
96 lrp_fsuid/lrp_fsgid. */
100 /* MDS has an authority for the Size-on-MDS attributes. */
101 LLIF_MDS_SIZE_LOCK = (1 << 0),
102 /* Epoch close is postponed. */
103 LLIF_EPOCH_PENDING = (1 << 1),
104 /* DONE WRITING is allowed. */
105 LLIF_DONE_WRITING = (1 << 2),
106 /* Sizeon-on-MDS attributes are changed. An attribute update needs to
108 LLIF_SOM_DIRTY = (1 << 3),
109 /* File is contented */
110 LLIF_CONTENDED = (1 << 4),
111 /* Truncate uses server lock for this file */
112 LLIF_SRVLOCK = (1 << 5)
116 struct ll_inode_info {
118 cfs_semaphore_t lli_size_sem; /* protect open and change size */
119 void *lli_size_sem_owner;
120 cfs_semaphore_t lli_write_sem;
121 cfs_rw_semaphore_t lli_trunc_sem;
122 char *lli_symlink_name;
125 unsigned long lli_flags;
126 cfs_time_t lli_contention_time;
128 /* this lock protects posix_acl, pending_write_llaps, mmap_cnt */
129 cfs_spinlock_t lli_lock;
130 cfs_list_t lli_close_list;
131 /* handle is to be sent to MDS later on done_writing and setattr.
132 * Open handle data are needed for the recovery to reconstruct
133 * the inode state on the MDS. XXX: recovery is not ready yet. */
134 struct obd_client_handle *lli_pending_och;
136 /* for writepage() only to communicate to fsync */
140 struct posix_acl *lli_posix_acl;
142 /* remote permission hash */
143 cfs_hlist_head_t *lli_remote_perms;
144 unsigned long lli_rmtperm_utime;
145 cfs_semaphore_t lli_rmtperm_sem;
147 cfs_list_t lli_dead_list;
149 cfs_semaphore_t lli_och_sem; /* Protects access to och pointers
150 and their usage counters, also
151 atomicity of check-update of
153 /* We need all three because every inode may be opened in different
155 struct obd_client_handle *lli_mds_read_och;
156 __u64 lli_open_fd_read_count;
157 struct obd_client_handle *lli_mds_write_och;
158 __u64 lli_open_fd_write_count;
159 struct obd_client_handle *lli_mds_exec_och;
160 __u64 lli_open_fd_exec_count;
162 struct inode lli_vfs_inode;
164 /* identifying fields for both metadata and data stacks. */
165 struct lu_fid lli_fid;
166 struct lov_stripe_md *lli_smd;
169 /* open count currently used by capability only, indicate whether
170 * capability needs renewal */
171 cfs_atomic_t lli_open_count;
172 struct obd_capa *lli_mds_capa;
173 cfs_list_t lli_oss_capas;
175 /* metadata statahead */
176 /* protect statahead stuff: lli_opendir_pid, lli_opendir_key, lli_sai,
178 cfs_spinlock_t lli_sa_lock;
180 * "opendir_pid" is the token when lookup/revalid -- I am the owner of
183 pid_t lli_opendir_pid;
185 * since parent-child threads can share the same @file struct,
186 * "opendir_key" is the token when dir close for case of parent exit
187 * before child -- it is me should cleanup the dir readahead. */
188 void *lli_opendir_key;
189 struct ll_statahead_info *lli_sai;
191 struct cl_object *lli_clob;
192 /* the most recent timestamps obtained from mds */
193 struct ost_lvb lli_lvb;
195 * serialize normal readdir and statahead-readdir
197 cfs_semaphore_t lli_readdir_sem;
201 * Locking to guarantee consistency of non-atomic updates to long long i_size,
202 * consistency between file size and KMS, and consistency within
203 * ->lli_smd->lsm_oinfo[]'s.
205 * Implemented by ->lli_size_sem and ->lsm_sem, nested in that order.
208 void ll_inode_size_lock(struct inode *inode, int lock_lsm);
209 void ll_inode_size_unlock(struct inode *inode, int unlock_lsm);
211 // FIXME: replace the name of this with LL_I to conform to kernel stuff
212 // static inline struct ll_inode_info *LL_I(struct inode *inode)
213 static inline struct ll_inode_info *ll_i2info(struct inode *inode)
215 return container_of(inode, struct ll_inode_info, lli_vfs_inode);
218 /* default to about 40meg of readahead on a given system. That much tied
219 * up in 512k readahead requests serviced at 40ms each is about 1GB/s. */
220 #define SBI_DEFAULT_READAHEAD_MAX (40UL << (20 - CFS_PAGE_SHIFT))
222 /* default to read-ahead full files smaller than 2MB on the second read */
223 #define SBI_DEFAULT_READAHEAD_WHOLE_MAX (2UL << (20 - CFS_PAGE_SHIFT))
228 RA_STAT_DISTANT_READPAGE,
229 RA_STAT_MISS_IN_WINDOW,
230 RA_STAT_FAILED_GRAB_PAGE,
231 RA_STAT_FAILED_MATCH,
236 RA_STAT_MAX_IN_FLIGHT,
237 RA_STAT_WRONG_GRAB_PAGE,
242 cfs_atomic_t ra_cur_pages;
243 unsigned long ra_max_pages;
244 unsigned long ra_max_pages_per_file;
245 unsigned long ra_max_read_ahead_whole_pages;
248 /* ra_io_arg will be filled in the beginning of ll_readahead with
249 * ras_lock, then the following ll_read_ahead_pages will read RA
250 * pages according to this arg, all the items in this structure are
251 * counted by page index.
254 unsigned long ria_start; /* start offset of read-ahead*/
255 unsigned long ria_end; /* end offset of read-ahead*/
256 /* If stride read pattern is detected, ria_stoff means where
257 * stride read is started. Note: for normal read-ahead, the
258 * value here is meaningless, and also it will not be accessed*/
260 /* ria_length and ria_pages are the length and pages length in the
261 * stride I/O mode. And they will also be used to check whether
262 * it is stride I/O read-ahead in the read-ahead pages*/
263 unsigned long ria_length;
264 unsigned long ria_pages;
267 /* LL_HIST_MAX=32 causes an overflow */
268 #define LL_HIST_MAX 28
269 #define LL_HIST_START 12 /* buckets start at 2^12 = 4k */
270 #define LL_PROCESS_HIST_MAX 10
271 struct per_process_info {
273 struct obd_histogram pp_r_hist;
274 struct obd_histogram pp_w_hist;
277 /* pp_extents[LL_PROCESS_HIST_MAX] will hold the combined process info */
278 struct ll_rw_extents_info {
279 struct per_process_info pp_extents[LL_PROCESS_HIST_MAX + 1];
282 #define LL_OFFSET_HIST_MAX 100
283 struct ll_rw_process_info {
286 loff_t rw_range_start;
288 loff_t rw_last_file_pos;
290 size_t rw_smallest_extent;
291 size_t rw_largest_extent;
292 struct ll_file_data *rw_last_file;
295 enum stats_track_type {
296 STATS_TRACK_ALL = 0, /* track all processes */
297 STATS_TRACK_PID, /* track process with this pid */
298 STATS_TRACK_PPID, /* track processes with this ppid */
299 STATS_TRACK_GID, /* track processes with this gid */
303 /* flags for sbi->ll_flags */
304 #define LL_SBI_NOLCK 0x01 /* DLM locking disabled (directio-only) */
305 #define LL_SBI_CHECKSUM 0x02 /* checksum each page as it's written */
306 #define LL_SBI_FLOCK 0x04
307 #define LL_SBI_USER_XATTR 0x08 /* support user xattr */
308 #define LL_SBI_ACL 0x10 /* support ACL */
309 #define LL_SBI_RMT_CLIENT 0x40 /* remote client */
310 #define LL_SBI_MDS_CAPA 0x80 /* support mds capa */
311 #define LL_SBI_OSS_CAPA 0x100 /* support oss capa */
312 #define LL_SBI_LOCALFLOCK 0x200 /* Local flocks support by kernel */
313 #define LL_SBI_LRU_RESIZE 0x400 /* lru resize support */
314 #define LL_SBI_LAZYSTATFS 0x800 /* lazystatfs mount option */
315 #define LL_SBI_SOM_PREVIEW 0x1000 /* SOM preview mount option */
316 #define LL_SBI_32BIT_API 0x2000 /* generate 32 bit inodes. */
318 /* default value for ll_sb_info->contention_time */
319 #define SBI_DEFAULT_CONTENTION_SECONDS 60
320 /* default value for lockless_truncate_enable */
321 #define SBI_DEFAULT_LOCKLESS_TRUNCATE_ENABLE 1
322 #define RCE_HASHES 32
324 struct rmtacl_ctl_entry {
326 pid_t rce_key; /* hash key */
327 int rce_ops; /* acl operation type */
330 struct rmtacl_ctl_table {
331 cfs_spinlock_t rct_lock;
332 cfs_list_t rct_entries[RCE_HASHES];
339 pid_t ee_key; /* hash key */
340 struct lu_fid ee_fid;
341 int ee_type; /* ACL type for ACCESS or DEFAULT */
342 ext_acl_xattr_header *ee_acl;
346 cfs_spinlock_t et_lock;
347 cfs_list_t et_entries[EE_HASHES];
352 /* this protects pglist and ra_info. It isn't safe to
353 * grab from interrupt contexts */
354 cfs_spinlock_t ll_lock;
355 cfs_spinlock_t ll_pp_extent_lock; /* Lock for pp_extent entries */
356 cfs_spinlock_t ll_process_lock; /* Lock for ll_rw_process_info */
357 struct obd_uuid ll_sb_uuid;
358 struct obd_export *ll_md_exp;
359 struct obd_export *ll_dt_exp;
360 struct proc_dir_entry* ll_proc_root;
361 struct lu_fid ll_root_fid; /* root object fid */
364 cfs_list_t ll_conn_chain; /* per-conn chain of SBs */
365 struct lustre_client_ocd ll_lco;
367 cfs_list_t ll_orphan_dentry_list; /*please don't ask -p*/
368 struct ll_close_queue *ll_lcq;
370 struct lprocfs_stats *ll_stats; /* lprocfs stats counter */
372 unsigned long ll_async_page_max;
373 unsigned long ll_async_page_count;
375 struct lprocfs_stats *ll_ra_stats;
377 struct ll_ra_info ll_ra_info;
378 unsigned int ll_namelen;
379 struct file_operations *ll_fop;
381 /* =0 - hold lock over whole read/write
382 * >0 - max. chunk to be read/written w/o lock re-acquiring */
383 unsigned long ll_max_rw_chunk;
385 struct lu_site *ll_site;
386 struct cl_device *ll_cl;
388 struct ll_rw_extents_info ll_rw_extents_info;
389 int ll_extent_process_count;
390 struct ll_rw_process_info ll_rw_process_info[LL_PROCESS_HIST_MAX];
391 unsigned int ll_offset_process_count;
392 struct ll_rw_process_info ll_rw_offset_info[LL_OFFSET_HIST_MAX];
393 unsigned int ll_rw_offset_entry_count;
394 enum stats_track_type ll_stats_track_type;
395 int ll_stats_track_id;
398 /* metadata stat-ahead */
399 unsigned int ll_sa_max; /* max statahead RPCs */
400 atomic_t ll_sa_total; /* statahead thread started
402 atomic_t ll_sa_wrong; /* statahead thread stopped for
405 dev_t ll_sdev_orig; /* save s_dev before assign for
407 struct rmtacl_ctl_table ll_rct;
408 struct eacl_table ll_et;
411 #define LL_DEFAULT_MAX_RW_CHUNK (32 * 1024 * 1024)
416 struct task_struct *lrr_reader;
417 cfs_list_t lrr_linkage;
421 * per file-descriptor read-ahead data.
423 struct ll_readahead_state {
424 cfs_spinlock_t ras_lock;
426 * index of the last page that read(2) needed and that wasn't in the
427 * cache. Used by ras_update() to detect seeks.
429 * XXX nikita: if access seeks into cached region, Lustre doesn't see
432 unsigned long ras_last_readpage;
434 * number of pages read after last read-ahead window reset. As window
435 * is reset on each seek, this is effectively a number of consecutive
436 * accesses. Maybe ->ras_accessed_in_window is better name.
438 * XXX nikita: window is also reset (by ras_update()) when Lustre
439 * believes that memory pressure evicts read-ahead pages. In that
440 * case, it probably doesn't make sense to expand window to
441 * PTLRPC_MAX_BRW_PAGES on the third access.
443 unsigned long ras_consecutive_pages;
445 * number of read requests after the last read-ahead window reset
446 * As window is reset on each seek, this is effectively the number
447 * on consecutive read request and is used to trigger read-ahead.
449 unsigned long ras_consecutive_requests;
451 * Parameters of current read-ahead window. Handled by
452 * ras_update(). On the initial access to the file or after a seek,
453 * window is reset to 0. After 3 consecutive accesses, window is
454 * expanded to PTLRPC_MAX_BRW_PAGES. Afterwards, window is enlarged by
455 * PTLRPC_MAX_BRW_PAGES chunks up to ->ra_max_pages.
457 unsigned long ras_window_start, ras_window_len;
459 * Where next read-ahead should start at. This lies within read-ahead
460 * window. Read-ahead window is read in pieces rather than at once
461 * because: 1. lustre limits total number of pages under read-ahead by
462 * ->ra_max_pages (see ll_ra_count_get()), 2. client cannot read pages
463 * not covered by DLM lock.
465 unsigned long ras_next_readahead;
467 * Total number of ll_file_read requests issued, reads originating
468 * due to mmap are not counted in this total. This value is used to
469 * trigger full file read-ahead after multiple reads to a small file.
471 unsigned long ras_requests;
473 * Page index with respect to the current request, these value
474 * will not be accurate when dealing with reads issued via mmap.
476 unsigned long ras_request_index;
478 * list of struct ll_ra_read's one per read(2) call current in
479 * progress against this file descriptor. Used by read-ahead code,
480 * protected by ->ras_lock.
482 cfs_list_t ras_read_beads;
484 * The following 3 items are used for detecting the stride I/O
486 * In stride I/O mode,
487 * ...............|-----data-----|****gap*****|--------|******|....
488 * offset |-stride_pages-|-stride_gap-|
489 * ras_stride_offset = offset;
490 * ras_stride_length = stride_pages + stride_gap;
491 * ras_stride_pages = stride_pages;
492 * Note: all these three items are counted by pages.
494 unsigned long ras_stride_length;
495 unsigned long ras_stride_pages;
496 pgoff_t ras_stride_offset;
498 * number of consecutive stride request count, and it is similar as
499 * ras_consecutive_requests, but used for stride I/O mode.
500 * Note: only more than 2 consecutive stride request are detected,
501 * stride read-ahead will be enable
503 unsigned long ras_consecutive_stride_requests;
511 extern cfs_mem_cache_t *ll_file_data_slab;
512 struct lustre_handle;
513 struct ll_file_data {
514 struct ll_readahead_state fd_ras;
516 struct ccc_grouplock fd_grouplock;
517 struct ll_file_dir fd_dir;
519 struct file *fd_file;
522 struct lov_stripe_md;
524 extern cfs_spinlock_t inode_lock;
526 extern struct proc_dir_entry *proc_lustre_fs_root;
528 static inline struct inode *ll_info2i(struct ll_inode_info *lli)
530 return &lli->lli_vfs_inode;
534 struct inode *icbd_parent;
535 struct dentry **icbd_childp;
539 __u32 ll_i2suppgid(struct inode *i);
540 void ll_i2gids(__u32 *suppgids, struct inode *i1,struct inode *i2);
542 static inline int ll_need_32bit_api(struct ll_sb_info *sbi)
544 #if BITS_PER_LONG == 32
547 return unlikely(cfs_curproc_is_32bit() || (sbi->ll_flags & LL_SBI_32BIT_API));
551 #define LLAP_MAGIC 98764321
553 extern cfs_mem_cache_t *ll_async_page_slab;
554 extern size_t ll_async_page_slab_size;
556 void ll_ra_read_in(struct file *f, struct ll_ra_read *rar);
557 void ll_ra_read_ex(struct file *f, struct ll_ra_read *rar);
558 struct ll_ra_read *ll_ra_read_get(struct file *f);
560 /* llite/lproc_llite.c */
562 int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
563 struct super_block *sb, char *osc, char *mdc);
564 void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi);
565 void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count);
566 void lprocfs_llite_init_vars(struct lprocfs_static_vars *lvars);
568 static inline int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
569 struct super_block *sb, char *osc, char *mdc){return 0;}
570 static inline void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi) {}
571 static void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count) {}
572 static void lprocfs_llite_init_vars(struct lprocfs_static_vars *lvars)
574 memset(lvars, 0, sizeof(*lvars));
580 static inline void ll_put_page(struct page *page)
583 page_cache_release(page);
586 extern struct file_operations ll_dir_operations;
587 extern struct inode_operations ll_dir_inode_operations;
588 struct page *ll_get_dir_page(struct file *filp, struct inode *dir, __u64 hash,
589 int exact, struct ll_dir_chain *chain);
590 int ll_readdir(struct file *filp, void *cookie, filldir_t filldir);
592 int ll_get_mdt_idx(struct inode *inode);
594 int ll_objects_destroy(struct ptlrpc_request *request,
596 struct inode *ll_iget(struct super_block *sb, ino_t hash,
597 struct lustre_md *lic);
598 int ll_md_blocking_ast(struct ldlm_lock *, struct ldlm_lock_desc *,
599 void *data, int flag);
600 struct lookup_intent *ll_convert_intent(struct open_intent *oit,
602 int ll_lookup_it_finish(struct ptlrpc_request *request,
603 struct lookup_intent *it, void *data);
606 int ll_prepare_write(struct file *, struct page *, unsigned from, unsigned to);
607 int ll_commit_write(struct file *, struct page *, unsigned from, unsigned to);
608 int ll_writepage(struct page *page, struct writeback_control *wbc);
609 void ll_removepage(struct page *page);
610 int ll_readpage(struct file *file, struct page *page);
611 void ll_readahead_init(struct inode *inode, struct ll_readahead_state *ras);
612 void ll_truncate(struct inode *inode);
613 int ll_file_punch(struct inode *, loff_t, int);
614 ssize_t ll_file_lockless_io(struct file *, char *, size_t, loff_t *, int);
615 void ll_clear_file_contended(struct inode*);
616 int ll_sync_page_range(struct inode *, struct address_space *, loff_t, size_t);
617 int ll_readahead(const struct lu_env *env, struct cl_io *io, struct ll_readahead_state *ras,
618 struct address_space *mapping, struct cl_page_list *queue, int flags);
621 extern struct file_operations ll_file_operations;
622 extern struct file_operations ll_file_operations_flock;
623 extern struct file_operations ll_file_operations_noflock;
624 extern struct inode_operations ll_file_inode_operations;
625 extern int ll_inode_revalidate_it(struct dentry *, struct lookup_intent *,
627 extern int ll_have_md_lock(struct inode *inode, __u64 bits, ldlm_mode_t l_req_mode);
628 extern ldlm_mode_t ll_take_md_lock(struct inode *inode, __u64 bits,
629 struct lustre_handle *lockh);
630 int __ll_inode_revalidate_it(struct dentry *, struct lookup_intent *, __u64 bits);
631 int ll_revalidate_nd(struct dentry *dentry, struct nameidata *nd);
632 int ll_file_open(struct inode *inode, struct file *file);
633 int ll_file_release(struct inode *inode, struct file *file);
634 int ll_glimpse_ioctl(struct ll_sb_info *sbi,
635 struct lov_stripe_md *lsm, lstat_t *st);
636 void ll_ioepoch_open(struct ll_inode_info *lli, __u64 ioepoch);
637 int ll_local_open(struct file *file,
638 struct lookup_intent *it, struct ll_file_data *fd,
639 struct obd_client_handle *och);
640 int ll_release_openhandle(struct dentry *, struct lookup_intent *);
641 int ll_md_close(struct obd_export *md_exp, struct inode *inode,
643 int ll_md_real_close(struct inode *inode, int flags);
644 void ll_ioepoch_close(struct inode *inode, struct md_op_data *op_data,
645 struct obd_client_handle **och, unsigned long flags);
646 void ll_done_writing_attr(struct inode *inode, struct md_op_data *op_data);
647 int ll_som_update(struct inode *inode, struct md_op_data *op_data);
648 int ll_inode_getattr(struct inode *inode, struct obdo *obdo,
649 __u64 ioepoch, int sync);
650 int ll_md_setattr(struct inode *inode, struct md_op_data *op_data,
651 struct md_open_data **mod);
652 void ll_pack_inode2opdata(struct inode *inode, struct md_op_data *op_data,
653 struct lustre_handle *fh);
654 extern void ll_rw_stats_tally(struct ll_sb_info *sbi, pid_t pid,
655 struct ll_file_data *file, loff_t pos,
656 size_t count, int rw);
657 int ll_getattr_it(struct vfsmount *mnt, struct dentry *de,
658 struct lookup_intent *it, struct kstat *stat);
659 int ll_getattr(struct vfsmount *mnt, struct dentry *de, struct kstat *stat);
660 struct ll_file_data *ll_file_data_get(void);
661 #ifndef HAVE_INODE_PERMISION_2ARGS
662 int ll_inode_permission(struct inode *inode, int mask, struct nameidata *nd);
664 int ll_inode_permission(struct inode *inode, int mask);
666 int ll_lov_setstripe_ea_info(struct inode *inode, struct file *file,
667 int flags, struct lov_user_md *lum,
669 int ll_lov_getstripe_ea_info(struct inode *inode, const char *filename,
670 struct lov_mds_md **lmm, int *lmm_size,
671 struct ptlrpc_request **request);
672 int ll_dir_setstripe(struct inode *inode, struct lov_user_md *lump,
674 int ll_dir_getstripe(struct inode *inode, struct lov_mds_md **lmm,
675 int *lmm_size, struct ptlrpc_request **request);
676 int ll_fsync(struct file *file, struct dentry *dentry, int data);
677 int ll_do_fiemap(struct inode *inode, struct ll_user_fiemap *fiemap,
679 int ll_merge_lvb(struct inode *inode);
680 int ll_get_grouplock(struct inode *inode, struct file *file, unsigned long arg);
681 int ll_put_grouplock(struct inode *inode, struct file *file, unsigned long arg);
682 int ll_fid2path(struct obd_export *exp, void *arg);
687 * protect race ll_find_aliases vs ll_revalidate_it vs ll_unhash_aliases
689 int ll_dops_init(struct dentry *de, int block, int init_sa);
690 extern cfs_spinlock_t ll_lookup_lock;
691 extern struct dentry_operations ll_d_ops;
692 void ll_intent_drop_lock(struct lookup_intent *);
693 void ll_intent_release(struct lookup_intent *);
694 int ll_drop_dentry(struct dentry *dentry);
695 int ll_drop_dentry(struct dentry *dentry);
696 void ll_unhash_aliases(struct inode *);
697 void ll_frob_intent(struct lookup_intent **itp, struct lookup_intent *deft);
698 void ll_lookup_finish_locks(struct lookup_intent *it, struct dentry *dentry);
699 int ll_dcompare(struct dentry *parent, struct qstr *d_name, struct qstr *name);
700 int ll_revalidate_it_finish(struct ptlrpc_request *request,
701 struct lookup_intent *it, struct dentry *de);
703 /* llite/llite_lib.c */
704 extern struct super_operations lustre_super_operations;
706 char *ll_read_opt(const char *opt, char *data);
707 void ll_lli_init(struct ll_inode_info *lli);
708 int ll_fill_super(struct super_block *sb);
709 void ll_put_super(struct super_block *sb);
710 void ll_kill_super(struct super_block *sb);
711 struct inode *ll_inode_from_lock(struct ldlm_lock *lock);
712 void ll_clear_inode(struct inode *inode);
713 int ll_setattr_raw(struct inode *inode, struct iattr *attr);
714 int ll_setattr(struct dentry *de, struct iattr *attr);
715 #ifndef HAVE_STATFS_DENTRY_PARAM
716 int ll_statfs(struct super_block *sb, struct kstatfs *sfs);
718 int ll_statfs(struct dentry *de, struct kstatfs *sfs);
720 int ll_statfs_internal(struct super_block *sb, struct obd_statfs *osfs,
721 __u64 max_age, __u32 flags);
722 void ll_update_inode(struct inode *inode, struct lustre_md *md);
723 void ll_read_inode2(struct inode *inode, void *opaque);
724 void ll_delete_inode(struct inode *inode);
725 int ll_iocontrol(struct inode *inode, struct file *file,
726 unsigned int cmd, unsigned long arg);
727 int ll_flush_ctx(struct inode *inode);
728 #ifdef HAVE_UMOUNTBEGIN_VFSMOUNT
729 void ll_umount_begin(struct vfsmount *vfsmnt, int flags);
731 void ll_umount_begin(struct super_block *sb);
733 int ll_remount_fs(struct super_block *sb, int *flags, char *data);
734 int ll_show_options(struct seq_file *seq, struct vfsmount *vfs);
735 int ll_prep_inode(struct inode **inode, struct ptlrpc_request *req,
736 struct super_block *);
737 void lustre_dump_dentry(struct dentry *, int recur);
738 void lustre_dump_inode(struct inode *);
739 int ll_obd_statfs(struct inode *inode, void *arg);
740 int ll_get_max_mdsize(struct ll_sb_info *sbi, int *max_mdsize);
741 int ll_process_config(struct lustre_cfg *lcfg);
742 struct md_op_data *ll_prep_md_op_data(struct md_op_data *op_data,
743 struct inode *i1, struct inode *i2,
744 const char *name, int namelen,
745 int mode, __u32 opc, void *data);
746 void ll_finish_md_op_data(struct md_op_data *op_data);
748 /* llite/llite_nfs.c */
749 extern struct export_operations lustre_export_operations;
750 __u32 get_uuid2int(const char *name, int len);
752 /* llite/special.c */
753 extern struct inode_operations ll_special_inode_operations;
754 extern struct file_operations ll_special_chr_inode_fops;
755 extern struct file_operations ll_special_chr_file_fops;
756 extern struct file_operations ll_special_blk_inode_fops;
757 extern struct file_operations ll_special_fifo_inode_fops;
758 extern struct file_operations ll_special_fifo_file_fops;
759 extern struct file_operations ll_special_sock_inode_fops;
761 /* llite/symlink.c */
762 extern struct inode_operations ll_fast_symlink_inode_operations;
764 /* llite/llite_close.c */
765 struct ll_close_queue {
766 cfs_spinlock_t lcq_lock;
768 cfs_waitq_t lcq_waitq;
769 cfs_completion_t lcq_comp;
770 cfs_atomic_t lcq_stop;
773 struct ccc_object *cl_inode2ccc(struct inode *inode);
776 void vvp_write_pending (struct ccc_object *club, struct ccc_page *page);
777 void vvp_write_complete(struct ccc_object *club, struct ccc_page *page);
779 /* specific achitecture can implement only part of this list */
780 enum vvp_io_subtype {
783 /** io called from .sendfile */
785 /** io started from splice_{read|write} */
792 enum vvp_io_subtype cui_io_subtype;
796 read_actor_t cui_actor;
800 struct pipe_inode_info *cui_pipe;
801 unsigned int cui_flags;
803 struct vvp_fault_io {
805 * Inode modification time that is checked across DLM
809 struct vm_area_struct *ft_vma;
811 * locked page returned from vvp_io
813 cfs_page_t *ft_vmpage;
814 #ifndef HAVE_VM_OP_FAULT
815 struct vm_nopage_api {
817 * Virtual address at which fault occurred.
819 unsigned long ft_address;
821 * Fault type, as to be supplied to
827 struct vm_fault_api {
831 struct vm_fault *ft_vmf;
833 * fault API used bitflags for return code.
835 unsigned int ft_flags;
841 * Read-ahead state used by read and page-fault IO contexts.
843 struct ll_ra_read cui_bead;
845 * Set when cui_bead has been initialized.
847 int cui_ra_window_set;
849 * Partially truncated page, that vvp_io_trunc_start() keeps locked
852 struct cl_page *cui_partpage;
856 * IO arguments for various VFS I/O interfaces.
859 /** normal/sendfile/splice */
860 enum vvp_io_subtype via_io_subtype;
864 #ifndef HAVE_FILE_WRITEV
865 struct kiocb *via_iocb;
867 struct iovec *via_iov;
868 unsigned long via_nrsegs;
871 read_actor_t via_actor;
875 struct pipe_inode_info *via_pipe;
876 unsigned int via_flags;
881 struct ll_cl_context {
883 struct cl_io *lcc_io;
884 struct cl_page *lcc_page;
885 struct lu_env *lcc_env;
890 struct vvp_thread_info {
891 struct ost_lvb vti_lvb;
892 struct cl_2queue vti_queue;
893 struct iovec vti_local_iov;
894 struct vvp_io_args vti_args;
895 struct ra_io_arg vti_ria;
896 struct kiocb vti_kiocb;
897 struct ll_cl_context vti_io_ctx;
900 static inline struct vvp_thread_info *vvp_env_info(const struct lu_env *env)
902 extern struct lu_context_key vvp_key;
903 struct vvp_thread_info *info;
905 info = lu_context_key_get(&env->le_ctx, &vvp_key);
906 LASSERT(info != NULL);
910 static inline struct vvp_io_args *vvp_env_args(const struct lu_env *env,
911 enum vvp_io_subtype type)
913 struct vvp_io_args *ret = &vvp_env_info(env)->vti_args;
915 ret->via_io_subtype = type;
921 struct vvp_io vs_ios;
924 static inline struct vvp_session *vvp_env_session(const struct lu_env *env)
926 extern struct lu_context_key vvp_session_key;
927 struct vvp_session *ses;
929 ses = lu_context_key_get(env->le_ses, &vvp_session_key);
930 LASSERT(ses != NULL);
934 static inline struct vvp_io *vvp_env_io(const struct lu_env *env)
936 return &vvp_env_session(env)->vs_ios;
939 void ll_queue_done_writing(struct inode *inode, unsigned long flags);
940 void ll_close_thread_shutdown(struct ll_close_queue *lcq);
941 int ll_close_thread_start(struct ll_close_queue **lcq_ret);
943 /* llite/llite_mmap.c */
944 typedef struct rb_root rb_root_t;
945 typedef struct rb_node rb_node_t;
947 struct ll_lock_tree_node;
948 struct ll_lock_tree {
950 cfs_list_t lt_locked_list;
951 struct ll_file_data *lt_fd;
954 int ll_teardown_mmaps(struct address_space *mapping, __u64 first, __u64 last);
955 int ll_file_mmap(struct file * file, struct vm_area_struct * vma);
956 struct ll_lock_tree_node * ll_node_from_inode(struct inode *inode, __u64 start,
957 __u64 end, ldlm_mode_t mode);
958 void policy_from_vma(ldlm_policy_data_t *policy,
959 struct vm_area_struct *vma, unsigned long addr, size_t count);
960 struct vm_area_struct *our_vma(unsigned long addr, size_t count);
962 #define ll_s2sbi(sb) (s2lsi(sb)->lsi_llsbi)
964 /* don't need an addref as the sb_info should be holding one */
965 static inline struct obd_export *ll_s2dtexp(struct super_block *sb)
967 return ll_s2sbi(sb)->ll_dt_exp;
970 /* don't need an addref as the sb_info should be holding one */
971 static inline struct obd_export *ll_s2mdexp(struct super_block *sb)
973 return ll_s2sbi(sb)->ll_md_exp;
976 static inline struct client_obd *sbi2mdc(struct ll_sb_info *sbi)
978 struct obd_device *obd = sbi->ll_md_exp->exp_obd;
984 // FIXME: replace the name of this with LL_SB to conform to kernel stuff
985 static inline struct ll_sb_info *ll_i2sbi(struct inode *inode)
987 return ll_s2sbi(inode->i_sb);
990 static inline struct obd_export *ll_i2dtexp(struct inode *inode)
992 return ll_s2dtexp(inode->i_sb);
995 static inline struct obd_export *ll_i2mdexp(struct inode *inode)
997 return ll_s2mdexp(inode->i_sb);
1000 static inline struct lu_fid *ll_inode2fid(struct inode *inode)
1003 LASSERT(inode != NULL);
1004 fid = &ll_i2info(inode)->lli_fid;
1005 LASSERT(fid_is_igif(fid) || fid_ver(fid) == 0);
1009 static inline int ll_mds_max_easize(struct super_block *sb)
1011 return sbi2mdc(ll_s2sbi(sb))->cl_max_mds_easize;
1014 static inline __u64 ll_file_maxbytes(struct inode *inode)
1016 return ll_i2info(inode)->lli_maxbytes;
1020 int ll_setxattr(struct dentry *dentry, const char *name,
1021 const void *value, size_t size, int flags);
1022 ssize_t ll_getxattr(struct dentry *dentry, const char *name,
1023 void *buffer, size_t size);
1024 ssize_t ll_listxattr(struct dentry *dentry, char *buffer, size_t size);
1025 int ll_removexattr(struct dentry *dentry, const char *name);
1027 /* llite/remote_perm.c */
1028 extern cfs_mem_cache_t *ll_remote_perm_cachep;
1029 extern cfs_mem_cache_t *ll_rmtperm_hash_cachep;
1031 cfs_hlist_head_t *alloc_rmtperm_hash(void);
1032 void free_rmtperm_hash(cfs_hlist_head_t *hash);
1033 int ll_update_remote_perm(struct inode *inode, struct mdt_remote_perm *perm);
1034 int lustre_check_remote_perm(struct inode *inode, int mask);
1036 /* llite/llite_capa.c */
1037 extern cfs_timer_t ll_capa_timer;
1039 int ll_capa_thread_start(void);
1040 void ll_capa_thread_stop(void);
1041 void ll_capa_timer_callback(unsigned long unused);
1043 struct obd_capa *ll_add_capa(struct inode *inode, struct obd_capa *ocapa);
1044 int ll_update_capa(struct obd_capa *ocapa, struct lustre_capa *capa);
1046 void ll_capa_open(struct inode *inode);
1047 void ll_capa_close(struct inode *inode);
1049 struct obd_capa *ll_mdscapa_get(struct inode *inode);
1050 struct obd_capa *ll_osscapa_get(struct inode *inode, __u64 opc);
1052 void ll_truncate_free_capa(struct obd_capa *ocapa);
1053 void ll_clear_inode_capas(struct inode *inode);
1054 void ll_print_capa_stat(struct ll_sb_info *sbi);
1056 /* llite/llite_cl.c */
1057 extern struct lu_device_type vvp_device_type;
1060 * Common IO arguments for various VFS I/O interfaces.
1063 int cl_sb_init(struct super_block *sb);
1064 int cl_sb_fini(struct super_block *sb);
1065 int cl_inode_init(struct inode *inode, struct lustre_md *md);
1066 void cl_inode_fini(struct inode *inode);
1068 enum cl_lock_mode vvp_mode_from_vma(struct vm_area_struct *vma);
1069 void ll_io_init(struct cl_io *io, const struct file *file, int write);
1071 void ras_update(struct ll_sb_info *sbi, struct inode *inode,
1072 struct ll_readahead_state *ras, unsigned long index,
1074 void ll_ra_count_put(struct ll_sb_info *sbi, unsigned long len);
1075 int ll_is_file_contended(struct file *file);
1076 void ll_ra_stats_inc(struct address_space *mapping, enum ra_stat which);
1078 /* llite/llite_rmtacl.c */
1079 #ifdef CONFIG_FS_POSIX_ACL
1080 obd_valid rce_ops2valid(int ops);
1081 struct rmtacl_ctl_entry *rct_search(struct rmtacl_ctl_table *rct, pid_t key);
1082 int rct_add(struct rmtacl_ctl_table *rct, pid_t key, int ops);
1083 int rct_del(struct rmtacl_ctl_table *rct, pid_t key);
1084 void rct_init(struct rmtacl_ctl_table *rct);
1085 void rct_fini(struct rmtacl_ctl_table *rct);
1087 void ee_free(struct eacl_entry *ee);
1088 int ee_add(struct eacl_table *et, pid_t key, struct lu_fid *fid, int type,
1089 ext_acl_xattr_header *header);
1090 struct eacl_entry *et_search_del(struct eacl_table *et, pid_t key,
1091 struct lu_fid *fid, int type);
1092 void et_search_free(struct eacl_table *et, pid_t key);
1093 void et_init(struct eacl_table *et);
1094 void et_fini(struct eacl_table *et);
1099 #define LL_SA_RPC_MIN 2
1100 #define LL_SA_RPC_DEF 32
1101 #define LL_SA_RPC_MAX 8192
1103 /* per inode struct, for dir only */
1104 struct ll_statahead_info {
1105 struct inode *sai_inode;
1106 unsigned int sai_generation; /* generation for statahead */
1107 cfs_atomic_t sai_refcount; /* when access this struct, hold
1109 unsigned int sai_sent; /* stat requests sent count */
1110 unsigned int sai_replied; /* stat requests which received
1112 unsigned int sai_max; /* max ahead of lookup */
1113 unsigned int sai_index; /* index of statahead entry */
1114 unsigned int sai_index_next; /* index for the next statahead
1115 * entry to be stated */
1116 unsigned int sai_hit; /* hit count */
1117 unsigned int sai_miss; /* miss count:
1118 * for "ls -al" case, it includes
1119 * hidden dentry miss;
1120 * for "ls -l" case, it does not
1121 * include hidden dentry miss.
1122 * "sai_miss_hidden" is used for
1125 unsigned int sai_consecutive_miss; /* consecutive miss */
1126 unsigned int sai_miss_hidden;/* "ls -al", but first dentry
1127 * is not a hidden one */
1128 unsigned int sai_skip_hidden;/* skipped hidden dentry count */
1129 unsigned int sai_ls_all:1; /* "ls -al", do stat-ahead for
1131 cfs_waitq_t sai_waitq; /* stat-ahead wait queue */
1132 struct ptlrpc_thread sai_thread; /* stat-ahead thread */
1133 cfs_list_t sai_entries_sent; /* entries sent out */
1134 cfs_list_t sai_entries_received; /* entries returned */
1135 cfs_list_t sai_entries_stated; /* entries stated */
1136 pid_t sai_pid; /* pid of statahead itself */
1139 int do_statahead_enter(struct inode *dir, struct dentry **dentry, int lookup);
1140 void ll_statahead_exit(struct inode *dir, struct dentry *dentry, int result);
1141 void ll_stop_statahead(struct inode *dir, void *key);
1144 void ll_statahead_mark(struct inode *dir, struct dentry *dentry)
1146 struct ll_inode_info *lli;
1147 struct ll_dentry_data *ldd = ll_d2d(dentry);
1149 /* dentry has been move to other directory, no need mark */
1150 if (unlikely(dir != dentry->d_parent->d_inode))
1153 lli = ll_i2info(dir);
1154 /* not the same process, don't mark */
1155 if (lli->lli_opendir_pid != cfs_curproc_pid())
1158 cfs_spin_lock(&lli->lli_sa_lock);
1159 if (likely(lli->lli_sai != NULL && ldd != NULL))
1160 ldd->lld_sa_generation = lli->lli_sai->sai_generation;
1161 cfs_spin_unlock(&lli->lli_sa_lock);
1165 int ll_statahead_enter(struct inode *dir, struct dentry **dentryp, int lookup)
1167 struct ll_inode_info *lli;
1168 struct ll_dentry_data *ldd = ll_d2d(*dentryp);
1170 if (unlikely(dir == NULL))
1173 if (ll_i2sbi(dir)->ll_sa_max == 0)
1176 lli = ll_i2info(dir);
1177 /* not the same process, don't statahead */
1178 if (lli->lli_opendir_pid != cfs_curproc_pid())
1182 * When "ls" a dentry, the system trigger more than once "revalidate" or
1183 * "lookup", for "getattr", for "getxattr", and maybe for others.
1184 * Under patchless client mode, the operation intent is not accurate,
1185 * it maybe misguide the statahead thread. For example:
1186 * The "revalidate" call for "getattr" and "getxattr" of a dentry maybe
1187 * have the same operation intent -- "IT_GETATTR".
1188 * In fact, one dentry should has only one chance to interact with the
1189 * statahead thread, otherwise the statahead windows will be confused.
1190 * The solution is as following:
1191 * Assign "lld_sa_generation" with "sai_generation" when a dentry
1192 * "IT_GETATTR" for the first time, and the subsequent "IT_GETATTR"
1193 * will bypass interacting with statahead thread for checking:
1194 * "lld_sa_generation == lli_sai->sai_generation"
1196 if (ldd && lli->lli_sai &&
1197 ldd->lld_sa_generation == lli->lli_sai->sai_generation)
1200 return do_statahead_enter(dir, dentryp, lookup);
1203 /* llite ioctl register support rountine */
1210 #define LLIOC_MAX_CMD 256
1213 * Rules to write a callback function:
1216 * @magic: Dynamic ioctl call routine will feed this vaule with the pointer
1217 * returned to ll_iocontrol_register. Callback functions should use this
1218 * data to check the potential collasion of ioctl cmd. If collasion is
1219 * found, callback function should return LLIOC_CONT.
1220 * @rcp: The result of ioctl command.
1223 * If @magic matches the pointer returned by ll_iocontrol_data, the
1224 * callback should return LLIOC_STOP; return LLIOC_STOP otherwise.
1226 typedef enum llioc_iter (*llioc_callback_t)(struct inode *inode,
1227 struct file *file, unsigned int cmd, unsigned long arg,
1228 void *magic, int *rcp);
1230 enum llioc_iter ll_iocontrol_call(struct inode *inode, struct file *file,
1231 unsigned int cmd, unsigned long arg, int *rcp);
1233 /* export functions */
1234 /* Register ioctl block dynamatically for a regular file.
1236 * @cmd: the array of ioctl command set
1237 * @count: number of commands in the @cmd
1238 * @cb: callback function, it will be called if an ioctl command is found to
1239 * belong to the command list @cmd.
1242 * A magic pointer will be returned if success;
1243 * otherwise, NULL will be returned.
1245 void *ll_iocontrol_register(llioc_callback_t cb, int count, unsigned int *cmd);
1246 void ll_iocontrol_unregister(void *magic);
1250 /* lclient compat stuff */
1251 #define cl_inode_info ll_inode_info
1252 #define cl_i2info(info) ll_i2info(info)
1253 #define cl_inode_mode(inode) ((inode)->i_mode)
1254 #define cl_i2sbi ll_i2sbi
1256 static inline void cl_isize_lock(struct inode *inode, int lsmlock)
1258 ll_inode_size_lock(inode, lsmlock);
1261 static inline void cl_isize_unlock(struct inode *inode, int lsmlock)
1263 ll_inode_size_unlock(inode, lsmlock);
1266 static inline void cl_isize_write_nolock(struct inode *inode, loff_t kms)
1268 LASSERT_SEM_LOCKED(&ll_i2info(inode)->lli_size_sem);
1269 i_size_write(inode, kms);
1272 static inline void cl_isize_write(struct inode *inode, loff_t kms)
1274 ll_inode_size_lock(inode, 0);
1275 i_size_write(inode, kms);
1276 ll_inode_size_unlock(inode, 0);
1279 #define cl_isize_read(inode) i_size_read(inode)
1281 static inline int cl_merge_lvb(struct inode *inode)
1283 return ll_merge_lvb(inode);
1286 #define cl_inode_atime(inode) LTIME_S((inode)->i_atime)
1287 #define cl_inode_ctime(inode) LTIME_S((inode)->i_ctime)
1288 #define cl_inode_mtime(inode) LTIME_S((inode)->i_mtime)
1290 struct obd_capa *cl_capa_lookup(struct inode *inode, enum cl_req_type crt);
1292 /** direct write pages */
1293 struct ll_dio_pages {
1294 /** page array to be written. we don't support
1295 * partial pages except the last one. */
1296 struct page **ldp_pages;
1297 /* offset of each page */
1298 loff_t *ldp_offsets;
1299 /** if ldp_offsets is NULL, it means a sequential
1300 * pages to be written, then this is the file offset
1301 * of the * first page. */
1302 loff_t ldp_start_offset;
1303 /** how many bytes are to be written. */
1305 /** # of pages in the array. */
1309 extern ssize_t ll_direct_rw_pages(const struct lu_env *env, struct cl_io *io,
1310 int rw, struct inode *inode,
1311 struct ll_dio_pages *pv);
1313 static inline int ll_file_nolock(const struct file *file)
1315 struct ll_file_data *fd = LUSTRE_FPRIVATE(file);
1316 struct inode *inode = file->f_dentry->d_inode;
1318 LASSERT(fd != NULL);
1319 return ((fd->fd_flags & LL_FILE_IGNORE_LOCK) ||
1320 (ll_i2sbi(inode)->ll_flags & LL_SBI_NOLCK));
1322 #endif /* LLITE_INTERNAL_H */