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 /** Only used on client-side for indicating the tail of dir hash/offset. */
56 #define LL_DIR_END_OFF 0x7fffffffffffffffULL
57 #define LL_DIR_END_OFF_32BIT 0x7fffffffUL
59 #ifndef DCACHE_LUSTRE_INVALID
60 #define DCACHE_LUSTRE_INVALID 0x4000000
63 #define LL_IT2STR(it) ((it) ? ldlm_it2str((it)->it_op) : "0")
64 #define LUSTRE_FPRIVATE(file) ((file)->private_data)
66 struct ll_dentry_data {
69 struct obd_client_handle lld_cwd_och;
70 struct obd_client_handle lld_mnt_och;
71 struct lookup_intent *lld_it;
72 unsigned int lld_sa_generation;
75 #define ll_d2d(de) ((struct ll_dentry_data*)((de)->d_fsdata))
77 extern struct file_operations ll_pgcache_seq_fops;
79 #define LLI_INODE_MAGIC 0x111d0de5
80 #define LLI_INODE_DEAD 0xdeadd00d
82 /* remote client permission cache */
83 #define REMOTE_PERM_HASHSIZE 16
85 struct ll_getname_data {
86 char *lgd_name; /* points to a buffer with NAME_MAX+1 size */
87 struct lu_fid lgd_fid; /* target fid we are looking for */
88 int lgd_found; /* inode matched? */
91 /* llite setxid/access permission for user on remote client */
92 struct ll_remote_perm {
93 cfs_hlist_node_t lrp_list;
98 int lrp_access_perm; /* MAY_READ/WRITE/EXEC, this
99 is access permission with
100 lrp_fsuid/lrp_fsgid. */
104 /* MDS has an authority for the Size-on-MDS attributes. */
105 LLIF_MDS_SIZE_LOCK = (1 << 0),
106 /* Epoch close is postponed. */
107 LLIF_EPOCH_PENDING = (1 << 1),
108 /* DONE WRITING is allowed. */
109 LLIF_DONE_WRITING = (1 << 2),
110 /* Sizeon-on-MDS attributes are changed. An attribute update needs to
112 LLIF_SOM_DIRTY = (1 << 3),
113 /* File is contented */
114 LLIF_CONTENDED = (1 << 4),
115 /* Truncate uses server lock for this file */
116 LLIF_SRVLOCK = (1 << 5)
120 struct ll_inode_info {
122 cfs_semaphore_t lli_size_sem; /* protect open and change size */
123 void *lli_size_sem_owner;
124 cfs_semaphore_t lli_write_sem;
125 cfs_rw_semaphore_t lli_trunc_sem;
126 char *lli_symlink_name;
129 unsigned long lli_flags;
130 cfs_time_t lli_contention_time;
132 /* this lock protects posix_acl, pending_write_llaps, mmap_cnt */
133 cfs_spinlock_t lli_lock;
134 cfs_list_t lli_close_list;
135 /* handle is to be sent to MDS later on done_writing and setattr.
136 * Open handle data are needed for the recovery to reconstruct
137 * the inode state on the MDS. XXX: recovery is not ready yet. */
138 struct obd_client_handle *lli_pending_och;
140 /* for writepage() only to communicate to fsync */
144 struct posix_acl *lli_posix_acl;
146 /* remote permission hash */
147 cfs_hlist_head_t *lli_remote_perms;
148 unsigned long lli_rmtperm_utime;
149 cfs_semaphore_t lli_rmtperm_sem;
151 cfs_list_t lli_dead_list;
153 cfs_semaphore_t lli_och_sem; /* Protects access to och pointers
154 and their usage counters, also
155 atomicity of check-update of
157 /* We need all three because every inode may be opened in different
159 struct obd_client_handle *lli_mds_read_och;
160 __u64 lli_open_fd_read_count;
161 struct obd_client_handle *lli_mds_write_och;
162 __u64 lli_open_fd_write_count;
163 struct obd_client_handle *lli_mds_exec_och;
164 __u64 lli_open_fd_exec_count;
166 struct inode lli_vfs_inode;
168 /* identifying fields for both metadata and data stacks. */
169 struct lu_fid lli_fid;
170 /* Parent fid for accessing default stripe data on parent directory
171 * for allocating OST objects after a mknod() and later open-by-FID. */
172 struct lu_fid lli_pfid;
173 struct lov_stripe_md *lli_smd;
176 /* open count currently used by capability only, indicate whether
177 * capability needs renewal */
178 cfs_atomic_t lli_open_count;
179 struct obd_capa *lli_mds_capa;
180 cfs_list_t lli_oss_capas;
182 /* metadata statahead */
183 /* protect statahead stuff: lli_opendir_pid, lli_opendir_key, lli_sai,
185 cfs_spinlock_t lli_sa_lock;
187 * "opendir_pid" is the token when lookup/revalid -- I am the owner of
190 pid_t lli_opendir_pid;
192 * since parent-child threads can share the same @file struct,
193 * "opendir_key" is the token when dir close for case of parent exit
194 * before child -- it is me should cleanup the dir readahead. */
195 void *lli_opendir_key;
196 struct ll_statahead_info *lli_sai;
198 struct cl_object *lli_clob;
199 /* the most recent timestamps obtained from mds */
200 struct ost_lvb lli_lvb;
202 * serialize normal readdir and statahead-readdir
204 cfs_semaphore_t lli_readdir_sem;
208 * Locking to guarantee consistency of non-atomic updates to long long i_size,
209 * consistency between file size and KMS, and consistency within
210 * ->lli_smd->lsm_oinfo[]'s.
212 * Implemented by ->lli_size_sem and ->lsm_sem, nested in that order.
215 void ll_inode_size_lock(struct inode *inode, int lock_lsm);
216 void ll_inode_size_unlock(struct inode *inode, int unlock_lsm);
218 // FIXME: replace the name of this with LL_I to conform to kernel stuff
219 // static inline struct ll_inode_info *LL_I(struct inode *inode)
220 static inline struct ll_inode_info *ll_i2info(struct inode *inode)
222 return container_of(inode, struct ll_inode_info, lli_vfs_inode);
225 /* default to about 40meg of readahead on a given system. That much tied
226 * up in 512k readahead requests serviced at 40ms each is about 1GB/s. */
227 #define SBI_DEFAULT_READAHEAD_MAX (40UL << (20 - CFS_PAGE_SHIFT))
229 /* default to read-ahead full files smaller than 2MB on the second read */
230 #define SBI_DEFAULT_READAHEAD_WHOLE_MAX (2UL << (20 - CFS_PAGE_SHIFT))
235 RA_STAT_DISTANT_READPAGE,
236 RA_STAT_MISS_IN_WINDOW,
237 RA_STAT_FAILED_GRAB_PAGE,
238 RA_STAT_FAILED_MATCH,
243 RA_STAT_MAX_IN_FLIGHT,
244 RA_STAT_WRONG_GRAB_PAGE,
249 cfs_atomic_t ra_cur_pages;
250 unsigned long ra_max_pages;
251 unsigned long ra_max_pages_per_file;
252 unsigned long ra_max_read_ahead_whole_pages;
255 /* ra_io_arg will be filled in the beginning of ll_readahead with
256 * ras_lock, then the following ll_read_ahead_pages will read RA
257 * pages according to this arg, all the items in this structure are
258 * counted by page index.
261 unsigned long ria_start; /* start offset of read-ahead*/
262 unsigned long ria_end; /* end offset of read-ahead*/
263 /* If stride read pattern is detected, ria_stoff means where
264 * stride read is started. Note: for normal read-ahead, the
265 * value here is meaningless, and also it will not be accessed*/
267 /* ria_length and ria_pages are the length and pages length in the
268 * stride I/O mode. And they will also be used to check whether
269 * it is stride I/O read-ahead in the read-ahead pages*/
270 unsigned long ria_length;
271 unsigned long ria_pages;
274 /* LL_HIST_MAX=32 causes an overflow */
275 #define LL_HIST_MAX 28
276 #define LL_HIST_START 12 /* buckets start at 2^12 = 4k */
277 #define LL_PROCESS_HIST_MAX 10
278 struct per_process_info {
280 struct obd_histogram pp_r_hist;
281 struct obd_histogram pp_w_hist;
284 /* pp_extents[LL_PROCESS_HIST_MAX] will hold the combined process info */
285 struct ll_rw_extents_info {
286 struct per_process_info pp_extents[LL_PROCESS_HIST_MAX + 1];
289 #define LL_OFFSET_HIST_MAX 100
290 struct ll_rw_process_info {
293 loff_t rw_range_start;
295 loff_t rw_last_file_pos;
297 size_t rw_smallest_extent;
298 size_t rw_largest_extent;
299 struct ll_file_data *rw_last_file;
302 enum stats_track_type {
303 STATS_TRACK_ALL = 0, /* track all processes */
304 STATS_TRACK_PID, /* track process with this pid */
305 STATS_TRACK_PPID, /* track processes with this ppid */
306 STATS_TRACK_GID, /* track processes with this gid */
310 /* flags for sbi->ll_flags */
311 #define LL_SBI_NOLCK 0x01 /* DLM locking disabled (directio-only) */
312 #define LL_SBI_CHECKSUM 0x02 /* checksum each page as it's written */
313 #define LL_SBI_FLOCK 0x04
314 #define LL_SBI_USER_XATTR 0x08 /* support user xattr */
315 #define LL_SBI_ACL 0x10 /* support ACL */
316 #define LL_SBI_RMT_CLIENT 0x40 /* remote client */
317 #define LL_SBI_MDS_CAPA 0x80 /* support mds capa */
318 #define LL_SBI_OSS_CAPA 0x100 /* support oss capa */
319 #define LL_SBI_LOCALFLOCK 0x200 /* Local flocks support by kernel */
320 #define LL_SBI_LRU_RESIZE 0x400 /* lru resize support */
321 #define LL_SBI_LAZYSTATFS 0x800 /* lazystatfs mount option */
322 #define LL_SBI_SOM_PREVIEW 0x1000 /* SOM preview mount option */
323 #define LL_SBI_32BIT_API 0x2000 /* generate 32 bit inodes. */
324 #define LL_SBI_64BIT_HASH 0x4000 /* support 64-bits dir hash/offset */
326 /* default value for ll_sb_info->contention_time */
327 #define SBI_DEFAULT_CONTENTION_SECONDS 60
328 /* default value for lockless_truncate_enable */
329 #define SBI_DEFAULT_LOCKLESS_TRUNCATE_ENABLE 1
330 #define RCE_HASHES 32
332 struct rmtacl_ctl_entry {
334 pid_t rce_key; /* hash key */
335 int rce_ops; /* acl operation type */
338 struct rmtacl_ctl_table {
339 cfs_spinlock_t rct_lock;
340 cfs_list_t rct_entries[RCE_HASHES];
347 pid_t ee_key; /* hash key */
348 struct lu_fid ee_fid;
349 int ee_type; /* ACL type for ACCESS or DEFAULT */
350 ext_acl_xattr_header *ee_acl;
354 cfs_spinlock_t et_lock;
355 cfs_list_t et_entries[EE_HASHES];
360 /* this protects pglist and ra_info. It isn't safe to
361 * grab from interrupt contexts */
362 cfs_spinlock_t ll_lock;
363 cfs_spinlock_t ll_pp_extent_lock; /* Lock for pp_extent entries */
364 cfs_spinlock_t ll_process_lock; /* Lock for ll_rw_process_info */
365 struct obd_uuid ll_sb_uuid;
366 struct obd_export *ll_md_exp;
367 struct obd_export *ll_dt_exp;
368 struct proc_dir_entry* ll_proc_root;
369 struct lu_fid ll_root_fid; /* root object fid */
372 cfs_list_t ll_conn_chain; /* per-conn chain of SBs */
373 struct lustre_client_ocd ll_lco;
375 cfs_list_t ll_orphan_dentry_list; /*please don't ask -p*/
376 struct ll_close_queue *ll_lcq;
378 struct lprocfs_stats *ll_stats; /* lprocfs stats counter */
380 unsigned long ll_async_page_max;
381 unsigned long ll_async_page_count;
383 struct lprocfs_stats *ll_ra_stats;
385 struct ll_ra_info ll_ra_info;
386 unsigned int ll_namelen;
387 struct file_operations *ll_fop;
389 /* =0 - hold lock over whole read/write
390 * >0 - max. chunk to be read/written w/o lock re-acquiring */
391 unsigned long ll_max_rw_chunk;
393 struct lu_site *ll_site;
394 struct cl_device *ll_cl;
396 struct ll_rw_extents_info ll_rw_extents_info;
397 int ll_extent_process_count;
398 struct ll_rw_process_info ll_rw_process_info[LL_PROCESS_HIST_MAX];
399 unsigned int ll_offset_process_count;
400 struct ll_rw_process_info ll_rw_offset_info[LL_OFFSET_HIST_MAX];
401 unsigned int ll_rw_offset_entry_count;
402 enum stats_track_type ll_stats_track_type;
403 int ll_stats_track_id;
406 /* metadata stat-ahead */
407 unsigned int ll_sa_max; /* max statahead RPCs */
408 atomic_t ll_sa_total; /* statahead thread started
410 atomic_t ll_sa_wrong; /* statahead thread stopped for
413 dev_t ll_sdev_orig; /* save s_dev before assign for
415 struct rmtacl_ctl_table ll_rct;
416 struct eacl_table ll_et;
417 struct vfsmount *ll_mnt;
420 #define LL_DEFAULT_MAX_RW_CHUNK (32 * 1024 * 1024)
425 struct task_struct *lrr_reader;
426 cfs_list_t lrr_linkage;
430 * per file-descriptor read-ahead data.
432 struct ll_readahead_state {
433 cfs_spinlock_t ras_lock;
435 * index of the last page that read(2) needed and that wasn't in the
436 * cache. Used by ras_update() to detect seeks.
438 * XXX nikita: if access seeks into cached region, Lustre doesn't see
441 unsigned long ras_last_readpage;
443 * number of pages read after last read-ahead window reset. As window
444 * is reset on each seek, this is effectively a number of consecutive
445 * accesses. Maybe ->ras_accessed_in_window is better name.
447 * XXX nikita: window is also reset (by ras_update()) when Lustre
448 * believes that memory pressure evicts read-ahead pages. In that
449 * case, it probably doesn't make sense to expand window to
450 * PTLRPC_MAX_BRW_PAGES on the third access.
452 unsigned long ras_consecutive_pages;
454 * number of read requests after the last read-ahead window reset
455 * As window is reset on each seek, this is effectively the number
456 * on consecutive read request and is used to trigger read-ahead.
458 unsigned long ras_consecutive_requests;
460 * Parameters of current read-ahead window. Handled by
461 * ras_update(). On the initial access to the file or after a seek,
462 * window is reset to 0. After 3 consecutive accesses, window is
463 * expanded to PTLRPC_MAX_BRW_PAGES. Afterwards, window is enlarged by
464 * PTLRPC_MAX_BRW_PAGES chunks up to ->ra_max_pages.
466 unsigned long ras_window_start, ras_window_len;
468 * Where next read-ahead should start at. This lies within read-ahead
469 * window. Read-ahead window is read in pieces rather than at once
470 * because: 1. lustre limits total number of pages under read-ahead by
471 * ->ra_max_pages (see ll_ra_count_get()), 2. client cannot read pages
472 * not covered by DLM lock.
474 unsigned long ras_next_readahead;
476 * Total number of ll_file_read requests issued, reads originating
477 * due to mmap are not counted in this total. This value is used to
478 * trigger full file read-ahead after multiple reads to a small file.
480 unsigned long ras_requests;
482 * Page index with respect to the current request, these value
483 * will not be accurate when dealing with reads issued via mmap.
485 unsigned long ras_request_index;
487 * list of struct ll_ra_read's one per read(2) call current in
488 * progress against this file descriptor. Used by read-ahead code,
489 * protected by ->ras_lock.
491 cfs_list_t ras_read_beads;
493 * The following 3 items are used for detecting the stride I/O
495 * In stride I/O mode,
496 * ...............|-----data-----|****gap*****|--------|******|....
497 * offset |-stride_pages-|-stride_gap-|
498 * ras_stride_offset = offset;
499 * ras_stride_length = stride_pages + stride_gap;
500 * ras_stride_pages = stride_pages;
501 * Note: all these three items are counted by pages.
503 unsigned long ras_stride_length;
504 unsigned long ras_stride_pages;
505 pgoff_t ras_stride_offset;
507 * number of consecutive stride request count, and it is similar as
508 * ras_consecutive_requests, but used for stride I/O mode.
509 * Note: only more than 2 consecutive stride request are detected,
510 * stride read-ahead will be enable
512 unsigned long ras_consecutive_stride_requests;
520 extern cfs_mem_cache_t *ll_file_data_slab;
521 struct lustre_handle;
522 struct ll_file_data {
523 struct ll_readahead_state fd_ras;
525 struct ccc_grouplock fd_grouplock;
526 struct ll_file_dir fd_dir;
528 struct file *fd_file;
531 struct lov_stripe_md;
533 extern cfs_spinlock_t inode_lock;
535 extern struct proc_dir_entry *proc_lustre_fs_root;
537 static inline struct inode *ll_info2i(struct ll_inode_info *lli)
539 return &lli->lli_vfs_inode;
543 struct inode *icbd_parent;
544 struct dentry **icbd_childp;
548 __u32 ll_i2suppgid(struct inode *i);
549 void ll_i2gids(__u32 *suppgids, struct inode *i1,struct inode *i2);
551 static inline int ll_need_32bit_api(struct ll_sb_info *sbi)
553 #if BITS_PER_LONG == 32
556 return unlikely(cfs_curproc_is_32bit() || (sbi->ll_flags & LL_SBI_32BIT_API));
560 #define LLAP_MAGIC 98764321
562 extern cfs_mem_cache_t *ll_async_page_slab;
563 extern size_t ll_async_page_slab_size;
565 void ll_ra_read_in(struct file *f, struct ll_ra_read *rar);
566 void ll_ra_read_ex(struct file *f, struct ll_ra_read *rar);
567 struct ll_ra_read *ll_ra_read_get(struct file *f);
569 /* llite/lproc_llite.c */
571 int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
572 struct super_block *sb, char *osc, char *mdc);
573 void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi);
574 void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count);
575 void lprocfs_llite_init_vars(struct lprocfs_static_vars *lvars);
577 static inline int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
578 struct super_block *sb, char *osc, char *mdc){return 0;}
579 static inline void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi) {}
580 static void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count) {}
581 static void lprocfs_llite_init_vars(struct lprocfs_static_vars *lvars)
583 memset(lvars, 0, sizeof(*lvars));
589 static inline void ll_put_page(struct page *page)
592 page_cache_release(page);
595 extern struct file_operations ll_dir_operations;
596 extern struct inode_operations ll_dir_inode_operations;
597 struct page *ll_get_dir_page(struct file *filp, struct inode *dir, __u64 hash,
598 int exact, struct ll_dir_chain *chain);
599 int ll_readdir(struct file *filp, void *cookie, filldir_t filldir);
601 int ll_get_mdt_idx(struct inode *inode);
603 int ll_objects_destroy(struct ptlrpc_request *request,
605 struct inode *ll_iget(struct super_block *sb, ino_t hash,
606 struct lustre_md *lic);
607 int ll_md_blocking_ast(struct ldlm_lock *, struct ldlm_lock_desc *,
608 void *data, int flag);
609 struct lookup_intent *ll_convert_intent(struct open_intent *oit,
611 int ll_lookup_it_finish(struct ptlrpc_request *request,
612 struct lookup_intent *it, void *data);
615 int ll_prepare_write(struct file *, struct page *, unsigned from, unsigned to);
616 int ll_commit_write(struct file *, struct page *, unsigned from, unsigned to);
617 int ll_writepage(struct page *page, struct writeback_control *wbc);
618 void ll_removepage(struct page *page);
619 int ll_readpage(struct file *file, struct page *page);
620 void ll_readahead_init(struct inode *inode, struct ll_readahead_state *ras);
621 void ll_truncate(struct inode *inode);
622 int ll_file_punch(struct inode *, loff_t, int);
623 ssize_t ll_file_lockless_io(struct file *, char *, size_t, loff_t *, int);
624 void ll_clear_file_contended(struct inode*);
625 int ll_sync_page_range(struct inode *, struct address_space *, loff_t, size_t);
626 int ll_readahead(const struct lu_env *env, struct cl_io *io, struct ll_readahead_state *ras,
627 struct address_space *mapping, struct cl_page_list *queue, int flags);
630 extern struct file_operations ll_file_operations;
631 extern struct file_operations ll_file_operations_flock;
632 extern struct file_operations ll_file_operations_noflock;
633 extern struct inode_operations ll_file_inode_operations;
634 extern int ll_inode_revalidate_it(struct dentry *, struct lookup_intent *,
636 extern int ll_have_md_lock(struct inode *inode, __u64 bits, ldlm_mode_t l_req_mode);
637 extern ldlm_mode_t ll_take_md_lock(struct inode *inode, __u64 bits,
638 struct lustre_handle *lockh);
639 int __ll_inode_revalidate_it(struct dentry *, struct lookup_intent *, __u64 bits);
640 int ll_revalidate_nd(struct dentry *dentry, struct nameidata *nd);
641 int ll_file_open(struct inode *inode, struct file *file);
642 int ll_file_release(struct inode *inode, struct file *file);
643 int ll_glimpse_ioctl(struct ll_sb_info *sbi,
644 struct lov_stripe_md *lsm, lstat_t *st);
645 void ll_ioepoch_open(struct ll_inode_info *lli, __u64 ioepoch);
646 int ll_local_open(struct file *file,
647 struct lookup_intent *it, struct ll_file_data *fd,
648 struct obd_client_handle *och);
649 int ll_release_openhandle(struct dentry *, struct lookup_intent *);
650 int ll_md_close(struct obd_export *md_exp, struct inode *inode,
652 int ll_md_real_close(struct inode *inode, int flags);
653 void ll_ioepoch_close(struct inode *inode, struct md_op_data *op_data,
654 struct obd_client_handle **och, unsigned long flags);
655 void ll_done_writing_attr(struct inode *inode, struct md_op_data *op_data);
656 int ll_som_update(struct inode *inode, struct md_op_data *op_data);
657 int ll_inode_getattr(struct inode *inode, struct obdo *obdo,
658 __u64 ioepoch, int sync);
659 int ll_md_setattr(struct inode *inode, struct md_op_data *op_data,
660 struct md_open_data **mod);
661 void ll_pack_inode2opdata(struct inode *inode, struct md_op_data *op_data,
662 struct lustre_handle *fh);
663 extern void ll_rw_stats_tally(struct ll_sb_info *sbi, pid_t pid,
664 struct ll_file_data *file, loff_t pos,
665 size_t count, int rw);
666 int ll_getattr_it(struct vfsmount *mnt, struct dentry *de,
667 struct lookup_intent *it, struct kstat *stat);
668 int ll_getattr(struct vfsmount *mnt, struct dentry *de, struct kstat *stat);
669 struct ll_file_data *ll_file_data_get(void);
670 #ifndef HAVE_INODE_PERMISION_2ARGS
671 int ll_inode_permission(struct inode *inode, int mask, struct nameidata *nd);
673 int ll_inode_permission(struct inode *inode, int mask);
675 int ll_lov_setstripe_ea_info(struct inode *inode, struct file *file,
676 int flags, struct lov_user_md *lum,
678 int ll_lov_getstripe_ea_info(struct inode *inode, const char *filename,
679 struct lov_mds_md **lmm, int *lmm_size,
680 struct ptlrpc_request **request);
681 int ll_dir_setstripe(struct inode *inode, struct lov_user_md *lump,
683 int ll_dir_getstripe(struct inode *inode, struct lov_mds_md **lmm,
684 int *lmm_size, struct ptlrpc_request **request);
685 int ll_fsync(struct file *file, struct dentry *dentry, int data);
686 int ll_do_fiemap(struct inode *inode, struct ll_user_fiemap *fiemap,
688 int ll_merge_lvb(struct inode *inode);
689 int ll_get_grouplock(struct inode *inode, struct file *file, unsigned long arg);
690 int ll_put_grouplock(struct inode *inode, struct file *file, unsigned long arg);
691 int ll_fid2path(struct obd_export *exp, void *arg);
696 * protect race ll_find_aliases vs ll_revalidate_it vs ll_unhash_aliases
698 int ll_dops_init(struct dentry *de, int block, int init_sa);
699 extern cfs_spinlock_t ll_lookup_lock;
700 extern struct dentry_operations ll_d_ops;
701 void ll_intent_drop_lock(struct lookup_intent *);
702 void ll_intent_release(struct lookup_intent *);
703 int ll_drop_dentry(struct dentry *dentry);
704 int ll_drop_dentry(struct dentry *dentry);
705 void ll_unhash_aliases(struct inode *);
706 void ll_frob_intent(struct lookup_intent **itp, struct lookup_intent *deft);
707 void ll_lookup_finish_locks(struct lookup_intent *it, struct dentry *dentry);
708 int ll_dcompare(struct dentry *parent, struct qstr *d_name, struct qstr *name);
709 int ll_revalidate_it_finish(struct ptlrpc_request *request,
710 struct lookup_intent *it, struct dentry *de);
712 /* llite/llite_lib.c */
713 extern struct super_operations lustre_super_operations;
715 char *ll_read_opt(const char *opt, char *data);
716 void ll_lli_init(struct ll_inode_info *lli);
717 int ll_fill_super(struct super_block *sb, struct vfsmount *mnt);
718 void ll_put_super(struct super_block *sb);
719 void ll_kill_super(struct super_block *sb);
720 struct inode *ll_inode_from_lock(struct ldlm_lock *lock);
721 void ll_clear_inode(struct inode *inode);
722 int ll_setattr_raw(struct inode *inode, struct iattr *attr);
723 int ll_setattr(struct dentry *de, struct iattr *attr);
724 #ifndef HAVE_STATFS_DENTRY_PARAM
725 int ll_statfs(struct super_block *sb, struct kstatfs *sfs);
727 int ll_statfs(struct dentry *de, struct kstatfs *sfs);
729 int ll_statfs_internal(struct super_block *sb, struct obd_statfs *osfs,
730 __u64 max_age, __u32 flags);
731 void ll_update_inode(struct inode *inode, struct lustre_md *md);
732 void ll_read_inode2(struct inode *inode, void *opaque);
733 void ll_delete_inode(struct inode *inode);
734 int ll_iocontrol(struct inode *inode, struct file *file,
735 unsigned int cmd, unsigned long arg);
736 int ll_flush_ctx(struct inode *inode);
737 #ifdef HAVE_UMOUNTBEGIN_VFSMOUNT
738 void ll_umount_begin(struct vfsmount *vfsmnt, int flags);
740 void ll_umount_begin(struct super_block *sb);
742 int ll_remount_fs(struct super_block *sb, int *flags, char *data);
743 int ll_show_options(struct seq_file *seq, struct vfsmount *vfs);
744 int ll_prep_inode(struct inode **inode, struct ptlrpc_request *req,
745 struct super_block *);
746 void lustre_dump_dentry(struct dentry *, int recur);
747 void lustre_dump_inode(struct inode *);
748 int ll_obd_statfs(struct inode *inode, void *arg);
749 int ll_get_max_mdsize(struct ll_sb_info *sbi, int *max_mdsize);
750 int ll_process_config(struct lustre_cfg *lcfg);
751 struct md_op_data *ll_prep_md_op_data(struct md_op_data *op_data,
752 struct inode *i1, struct inode *i2,
753 const char *name, int namelen,
754 int mode, __u32 opc, void *data);
755 void ll_finish_md_op_data(struct md_op_data *op_data);
757 /* llite/llite_nfs.c */
758 extern struct export_operations lustre_export_operations;
759 __u32 get_uuid2int(const char *name, int len);
761 /* llite/special.c */
762 extern struct inode_operations ll_special_inode_operations;
763 extern struct file_operations ll_special_chr_inode_fops;
764 extern struct file_operations ll_special_chr_file_fops;
765 extern struct file_operations ll_special_blk_inode_fops;
766 extern struct file_operations ll_special_fifo_inode_fops;
767 extern struct file_operations ll_special_fifo_file_fops;
768 extern struct file_operations ll_special_sock_inode_fops;
770 /* llite/symlink.c */
771 extern struct inode_operations ll_fast_symlink_inode_operations;
773 /* llite/llite_close.c */
774 struct ll_close_queue {
775 cfs_spinlock_t lcq_lock;
777 cfs_waitq_t lcq_waitq;
778 cfs_completion_t lcq_comp;
779 cfs_atomic_t lcq_stop;
782 struct ccc_object *cl_inode2ccc(struct inode *inode);
785 void vvp_write_pending (struct ccc_object *club, struct ccc_page *page);
786 void vvp_write_complete(struct ccc_object *club, struct ccc_page *page);
788 /* specific achitecture can implement only part of this list */
789 enum vvp_io_subtype {
792 /** io called from .sendfile */
794 /** io started from splice_{read|write} */
801 enum vvp_io_subtype cui_io_subtype;
805 read_actor_t cui_actor;
809 struct pipe_inode_info *cui_pipe;
810 unsigned int cui_flags;
812 struct vvp_fault_io {
814 * Inode modification time that is checked across DLM
818 struct vm_area_struct *ft_vma;
820 * locked page returned from vvp_io
822 cfs_page_t *ft_vmpage;
823 #ifndef HAVE_VM_OP_FAULT
824 struct vm_nopage_api {
826 * Virtual address at which fault occurred.
828 unsigned long ft_address;
830 * Fault type, as to be supplied to
836 struct vm_fault_api {
840 struct vm_fault *ft_vmf;
842 * fault API used bitflags for return code.
844 unsigned int ft_flags;
850 * Read-ahead state used by read and page-fault IO contexts.
852 struct ll_ra_read cui_bead;
854 * Set when cui_bead has been initialized.
856 int cui_ra_window_set;
858 * Partially truncated page, that vvp_io_trunc_start() keeps locked
861 struct cl_page *cui_partpage;
865 * IO arguments for various VFS I/O interfaces.
868 /** normal/sendfile/splice */
869 enum vvp_io_subtype via_io_subtype;
873 #ifndef HAVE_FILE_WRITEV
874 struct kiocb *via_iocb;
876 struct iovec *via_iov;
877 unsigned long via_nrsegs;
880 read_actor_t via_actor;
884 struct pipe_inode_info *via_pipe;
885 unsigned int via_flags;
890 struct ll_cl_context {
892 struct cl_io *lcc_io;
893 struct cl_page *lcc_page;
894 struct lu_env *lcc_env;
899 struct vvp_thread_info {
900 struct ost_lvb vti_lvb;
901 struct cl_2queue vti_queue;
902 struct iovec vti_local_iov;
903 struct vvp_io_args vti_args;
904 struct ra_io_arg vti_ria;
905 struct kiocb vti_kiocb;
906 struct ll_cl_context vti_io_ctx;
909 static inline struct vvp_thread_info *vvp_env_info(const struct lu_env *env)
911 extern struct lu_context_key vvp_key;
912 struct vvp_thread_info *info;
914 info = lu_context_key_get(&env->le_ctx, &vvp_key);
915 LASSERT(info != NULL);
919 static inline struct vvp_io_args *vvp_env_args(const struct lu_env *env,
920 enum vvp_io_subtype type)
922 struct vvp_io_args *ret = &vvp_env_info(env)->vti_args;
924 ret->via_io_subtype = type;
930 struct vvp_io vs_ios;
933 static inline struct vvp_session *vvp_env_session(const struct lu_env *env)
935 extern struct lu_context_key vvp_session_key;
936 struct vvp_session *ses;
938 ses = lu_context_key_get(env->le_ses, &vvp_session_key);
939 LASSERT(ses != NULL);
943 static inline struct vvp_io *vvp_env_io(const struct lu_env *env)
945 return &vvp_env_session(env)->vs_ios;
948 void ll_queue_done_writing(struct inode *inode, unsigned long flags);
949 void ll_close_thread_shutdown(struct ll_close_queue *lcq);
950 int ll_close_thread_start(struct ll_close_queue **lcq_ret);
952 /* llite/llite_mmap.c */
953 typedef struct rb_root rb_root_t;
954 typedef struct rb_node rb_node_t;
956 struct ll_lock_tree_node;
957 struct ll_lock_tree {
959 cfs_list_t lt_locked_list;
960 struct ll_file_data *lt_fd;
963 int ll_teardown_mmaps(struct address_space *mapping, __u64 first, __u64 last);
964 int ll_file_mmap(struct file * file, struct vm_area_struct * vma);
965 struct ll_lock_tree_node * ll_node_from_inode(struct inode *inode, __u64 start,
966 __u64 end, ldlm_mode_t mode);
967 void policy_from_vma(ldlm_policy_data_t *policy,
968 struct vm_area_struct *vma, unsigned long addr, size_t count);
969 struct vm_area_struct *our_vma(unsigned long addr, size_t count);
971 #define ll_s2sbi(sb) (s2lsi(sb)->lsi_llsbi)
973 /* don't need an addref as the sb_info should be holding one */
974 static inline struct obd_export *ll_s2dtexp(struct super_block *sb)
976 return ll_s2sbi(sb)->ll_dt_exp;
979 /* don't need an addref as the sb_info should be holding one */
980 static inline struct obd_export *ll_s2mdexp(struct super_block *sb)
982 return ll_s2sbi(sb)->ll_md_exp;
985 static inline struct client_obd *sbi2mdc(struct ll_sb_info *sbi)
987 struct obd_device *obd = sbi->ll_md_exp->exp_obd;
993 // FIXME: replace the name of this with LL_SB to conform to kernel stuff
994 static inline struct ll_sb_info *ll_i2sbi(struct inode *inode)
996 return ll_s2sbi(inode->i_sb);
999 static inline struct obd_export *ll_i2dtexp(struct inode *inode)
1001 return ll_s2dtexp(inode->i_sb);
1004 static inline struct obd_export *ll_i2mdexp(struct inode *inode)
1006 return ll_s2mdexp(inode->i_sb);
1009 static inline struct lu_fid *ll_inode2fid(struct inode *inode)
1012 LASSERT(inode != NULL);
1013 fid = &ll_i2info(inode)->lli_fid;
1014 LASSERT(fid_is_igif(fid) || fid_ver(fid) == 0);
1018 static inline int ll_mds_max_easize(struct super_block *sb)
1020 return sbi2mdc(ll_s2sbi(sb))->cl_max_mds_easize;
1023 static inline __u64 ll_file_maxbytes(struct inode *inode)
1025 return ll_i2info(inode)->lli_maxbytes;
1029 int ll_setxattr(struct dentry *dentry, const char *name,
1030 const void *value, size_t size, int flags);
1031 ssize_t ll_getxattr(struct dentry *dentry, const char *name,
1032 void *buffer, size_t size);
1033 ssize_t ll_listxattr(struct dentry *dentry, char *buffer, size_t size);
1034 int ll_removexattr(struct dentry *dentry, const char *name);
1036 /* llite/remote_perm.c */
1037 extern cfs_mem_cache_t *ll_remote_perm_cachep;
1038 extern cfs_mem_cache_t *ll_rmtperm_hash_cachep;
1040 cfs_hlist_head_t *alloc_rmtperm_hash(void);
1041 void free_rmtperm_hash(cfs_hlist_head_t *hash);
1042 int ll_update_remote_perm(struct inode *inode, struct mdt_remote_perm *perm);
1043 int lustre_check_remote_perm(struct inode *inode, int mask);
1045 /* llite/llite_capa.c */
1046 extern cfs_timer_t ll_capa_timer;
1048 int ll_capa_thread_start(void);
1049 void ll_capa_thread_stop(void);
1050 void ll_capa_timer_callback(unsigned long unused);
1052 struct obd_capa *ll_add_capa(struct inode *inode, struct obd_capa *ocapa);
1053 int ll_update_capa(struct obd_capa *ocapa, struct lustre_capa *capa);
1055 void ll_capa_open(struct inode *inode);
1056 void ll_capa_close(struct inode *inode);
1058 struct obd_capa *ll_mdscapa_get(struct inode *inode);
1059 struct obd_capa *ll_osscapa_get(struct inode *inode, __u64 opc);
1061 void ll_truncate_free_capa(struct obd_capa *ocapa);
1062 void ll_clear_inode_capas(struct inode *inode);
1063 void ll_print_capa_stat(struct ll_sb_info *sbi);
1065 /* llite/llite_cl.c */
1066 extern struct lu_device_type vvp_device_type;
1069 * Common IO arguments for various VFS I/O interfaces.
1072 int cl_sb_init(struct super_block *sb);
1073 int cl_sb_fini(struct super_block *sb);
1074 int cl_inode_init(struct inode *inode, struct lustre_md *md);
1075 void cl_inode_fini(struct inode *inode);
1077 enum cl_lock_mode vvp_mode_from_vma(struct vm_area_struct *vma);
1078 void ll_io_init(struct cl_io *io, const struct file *file, int write);
1080 void ras_update(struct ll_sb_info *sbi, struct inode *inode,
1081 struct ll_readahead_state *ras, unsigned long index,
1083 void ll_ra_count_put(struct ll_sb_info *sbi, unsigned long len);
1084 int ll_is_file_contended(struct file *file);
1085 void ll_ra_stats_inc(struct address_space *mapping, enum ra_stat which);
1087 /* llite/llite_rmtacl.c */
1088 #ifdef CONFIG_FS_POSIX_ACL
1089 obd_valid rce_ops2valid(int ops);
1090 struct rmtacl_ctl_entry *rct_search(struct rmtacl_ctl_table *rct, pid_t key);
1091 int rct_add(struct rmtacl_ctl_table *rct, pid_t key, int ops);
1092 int rct_del(struct rmtacl_ctl_table *rct, pid_t key);
1093 void rct_init(struct rmtacl_ctl_table *rct);
1094 void rct_fini(struct rmtacl_ctl_table *rct);
1096 void ee_free(struct eacl_entry *ee);
1097 int ee_add(struct eacl_table *et, pid_t key, struct lu_fid *fid, int type,
1098 ext_acl_xattr_header *header);
1099 struct eacl_entry *et_search_del(struct eacl_table *et, pid_t key,
1100 struct lu_fid *fid, int type);
1101 void et_search_free(struct eacl_table *et, pid_t key);
1102 void et_init(struct eacl_table *et);
1103 void et_fini(struct eacl_table *et);
1108 #define LL_SA_RPC_MIN 2
1109 #define LL_SA_RPC_DEF 32
1110 #define LL_SA_RPC_MAX 8192
1112 /* per inode struct, for dir only */
1113 struct ll_statahead_info {
1114 struct inode *sai_inode;
1115 unsigned int sai_generation; /* generation for statahead */
1116 cfs_atomic_t sai_refcount; /* when access this struct, hold
1118 unsigned int sai_sent; /* stat requests sent count */
1119 unsigned int sai_replied; /* stat requests which received
1121 unsigned int sai_max; /* max ahead of lookup */
1122 unsigned int sai_index; /* index of statahead entry */
1123 unsigned int sai_index_next; /* index for the next statahead
1124 * entry to be stated */
1125 unsigned int sai_hit; /* hit count */
1126 unsigned int sai_miss; /* miss count:
1127 * for "ls -al" case, it includes
1128 * hidden dentry miss;
1129 * for "ls -l" case, it does not
1130 * include hidden dentry miss.
1131 * "sai_miss_hidden" is used for
1134 unsigned int sai_consecutive_miss; /* consecutive miss */
1135 unsigned int sai_miss_hidden;/* "ls -al", but first dentry
1136 * is not a hidden one */
1137 unsigned int sai_skip_hidden;/* skipped hidden dentry count */
1138 unsigned int sai_ls_all:1; /* "ls -al", do stat-ahead for
1140 cfs_waitq_t sai_waitq; /* stat-ahead wait queue */
1141 struct ptlrpc_thread sai_thread; /* stat-ahead thread */
1142 cfs_list_t sai_entries_sent; /* entries sent out */
1143 cfs_list_t sai_entries_received; /* entries returned */
1144 cfs_list_t sai_entries_stated; /* entries stated */
1145 pid_t sai_pid; /* pid of statahead itself */
1148 int do_statahead_enter(struct inode *dir, struct dentry **dentry, int lookup);
1149 void ll_statahead_exit(struct inode *dir, struct dentry *dentry, int result);
1150 void ll_stop_statahead(struct inode *dir, void *key);
1153 void ll_statahead_mark(struct inode *dir, struct dentry *dentry)
1155 struct ll_inode_info *lli;
1156 struct ll_dentry_data *ldd = ll_d2d(dentry);
1158 /* dentry has been move to other directory, no need mark */
1159 if (unlikely(dir != dentry->d_parent->d_inode))
1162 lli = ll_i2info(dir);
1163 /* not the same process, don't mark */
1164 if (lli->lli_opendir_pid != cfs_curproc_pid())
1167 cfs_spin_lock(&lli->lli_sa_lock);
1168 if (likely(lli->lli_sai != NULL && ldd != NULL))
1169 ldd->lld_sa_generation = lli->lli_sai->sai_generation;
1170 cfs_spin_unlock(&lli->lli_sa_lock);
1174 int ll_statahead_enter(struct inode *dir, struct dentry **dentryp, int lookup)
1176 struct ll_inode_info *lli;
1177 struct ll_dentry_data *ldd = ll_d2d(*dentryp);
1179 if (unlikely(dir == NULL))
1182 if (ll_i2sbi(dir)->ll_sa_max == 0)
1185 lli = ll_i2info(dir);
1186 /* not the same process, don't statahead */
1187 if (lli->lli_opendir_pid != cfs_curproc_pid())
1191 * When "ls" a dentry, the system trigger more than once "revalidate" or
1192 * "lookup", for "getattr", for "getxattr", and maybe for others.
1193 * Under patchless client mode, the operation intent is not accurate,
1194 * it maybe misguide the statahead thread. For example:
1195 * The "revalidate" call for "getattr" and "getxattr" of a dentry maybe
1196 * have the same operation intent -- "IT_GETATTR".
1197 * In fact, one dentry should has only one chance to interact with the
1198 * statahead thread, otherwise the statahead windows will be confused.
1199 * The solution is as following:
1200 * Assign "lld_sa_generation" with "sai_generation" when a dentry
1201 * "IT_GETATTR" for the first time, and the subsequent "IT_GETATTR"
1202 * will bypass interacting with statahead thread for checking:
1203 * "lld_sa_generation == lli_sai->sai_generation"
1205 if (ldd && lli->lli_sai &&
1206 ldd->lld_sa_generation == lli->lli_sai->sai_generation)
1209 return do_statahead_enter(dir, dentryp, lookup);
1212 /* llite ioctl register support rountine */
1219 #define LLIOC_MAX_CMD 256
1222 * Rules to write a callback function:
1225 * @magic: Dynamic ioctl call routine will feed this vaule with the pointer
1226 * returned to ll_iocontrol_register. Callback functions should use this
1227 * data to check the potential collasion of ioctl cmd. If collasion is
1228 * found, callback function should return LLIOC_CONT.
1229 * @rcp: The result of ioctl command.
1232 * If @magic matches the pointer returned by ll_iocontrol_data, the
1233 * callback should return LLIOC_STOP; return LLIOC_STOP otherwise.
1235 typedef enum llioc_iter (*llioc_callback_t)(struct inode *inode,
1236 struct file *file, unsigned int cmd, unsigned long arg,
1237 void *magic, int *rcp);
1239 enum llioc_iter ll_iocontrol_call(struct inode *inode, struct file *file,
1240 unsigned int cmd, unsigned long arg, int *rcp);
1242 /* export functions */
1243 /* Register ioctl block dynamatically for a regular file.
1245 * @cmd: the array of ioctl command set
1246 * @count: number of commands in the @cmd
1247 * @cb: callback function, it will be called if an ioctl command is found to
1248 * belong to the command list @cmd.
1251 * A magic pointer will be returned if success;
1252 * otherwise, NULL will be returned.
1254 void *ll_iocontrol_register(llioc_callback_t cb, int count, unsigned int *cmd);
1255 void ll_iocontrol_unregister(void *magic);
1259 /* lclient compat stuff */
1260 #define cl_inode_info ll_inode_info
1261 #define cl_i2info(info) ll_i2info(info)
1262 #define cl_inode_mode(inode) ((inode)->i_mode)
1263 #define cl_i2sbi ll_i2sbi
1265 static inline void cl_isize_lock(struct inode *inode, int lsmlock)
1267 ll_inode_size_lock(inode, lsmlock);
1270 static inline void cl_isize_unlock(struct inode *inode, int lsmlock)
1272 ll_inode_size_unlock(inode, lsmlock);
1275 static inline void cl_isize_write_nolock(struct inode *inode, loff_t kms)
1277 LASSERT_SEM_LOCKED(&ll_i2info(inode)->lli_size_sem);
1278 i_size_write(inode, kms);
1281 static inline void cl_isize_write(struct inode *inode, loff_t kms)
1283 ll_inode_size_lock(inode, 0);
1284 i_size_write(inode, kms);
1285 ll_inode_size_unlock(inode, 0);
1288 #define cl_isize_read(inode) i_size_read(inode)
1290 static inline int cl_merge_lvb(struct inode *inode)
1292 return ll_merge_lvb(inode);
1295 #define cl_inode_atime(inode) LTIME_S((inode)->i_atime)
1296 #define cl_inode_ctime(inode) LTIME_S((inode)->i_ctime)
1297 #define cl_inode_mtime(inode) LTIME_S((inode)->i_mtime)
1299 struct obd_capa *cl_capa_lookup(struct inode *inode, enum cl_req_type crt);
1301 /** direct write pages */
1302 struct ll_dio_pages {
1303 /** page array to be written. we don't support
1304 * partial pages except the last one. */
1305 struct page **ldp_pages;
1306 /* offset of each page */
1307 loff_t *ldp_offsets;
1308 /** if ldp_offsets is NULL, it means a sequential
1309 * pages to be written, then this is the file offset
1310 * of the * first page. */
1311 loff_t ldp_start_offset;
1312 /** how many bytes are to be written. */
1314 /** # of pages in the array. */
1318 extern ssize_t ll_direct_rw_pages(const struct lu_env *env, struct cl_io *io,
1319 int rw, struct inode *inode,
1320 struct ll_dio_pages *pv);
1322 static inline int ll_file_nolock(const struct file *file)
1324 struct ll_file_data *fd = LUSTRE_FPRIVATE(file);
1325 struct inode *inode = file->f_dentry->d_inode;
1327 LASSERT(fd != NULL);
1328 return ((fd->fd_flags & LL_FILE_IGNORE_LOCK) ||
1329 (ll_i2sbi(inode)->ll_flags & LL_SBI_NOLCK));
1331 #endif /* LLITE_INTERNAL_H */