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 VM_FAULT_RETRY
56 #define VM_FAULT_RETRY 0
59 /** Only used on client-side for indicating the tail of dir hash/offset. */
60 #define LL_DIR_END_OFF 0x7fffffffffffffffULL
61 #define LL_DIR_END_OFF_32BIT 0x7fffffffUL
63 #ifndef DCACHE_LUSTRE_INVALID
64 #define DCACHE_LUSTRE_INVALID 0x4000000
67 #define LL_IT2STR(it) ((it) ? ldlm_it2str((it)->it_op) : "0")
68 #define LUSTRE_FPRIVATE(file) ((file)->private_data)
70 struct ll_dentry_data {
73 struct obd_client_handle lld_cwd_och;
74 struct obd_client_handle lld_mnt_och;
75 struct lookup_intent *lld_it;
76 unsigned int lld_sa_generation;
79 #define ll_d2d(de) ((struct ll_dentry_data*)((de)->d_fsdata))
81 extern struct file_operations ll_pgcache_seq_fops;
83 #define LLI_INODE_MAGIC 0x111d0de5
84 #define LLI_INODE_DEAD 0xdeadd00d
86 /* remote client permission cache */
87 #define REMOTE_PERM_HASHSIZE 16
89 struct ll_getname_data {
90 char *lgd_name; /* points to a buffer with NAME_MAX+1 size */
91 struct lu_fid lgd_fid; /* target fid we are looking for */
92 int lgd_found; /* inode matched? */
95 /* llite setxid/access permission for user on remote client */
96 struct ll_remote_perm {
97 cfs_hlist_node_t lrp_list;
102 int lrp_access_perm; /* MAY_READ/WRITE/EXEC, this
103 is access permission with
104 lrp_fsuid/lrp_fsgid. */
108 /* MDS has an authority for the Size-on-MDS attributes. */
109 LLIF_MDS_SIZE_LOCK = (1 << 0),
110 /* Epoch close is postponed. */
111 LLIF_EPOCH_PENDING = (1 << 1),
112 /* DONE WRITING is allowed. */
113 LLIF_DONE_WRITING = (1 << 2),
114 /* Sizeon-on-MDS attributes are changed. An attribute update needs to
116 LLIF_SOM_DIRTY = (1 << 3),
117 /* File is contented */
118 LLIF_CONTENDED = (1 << 4),
119 /* Truncate uses server lock for this file */
120 LLIF_SRVLOCK = (1 << 5)
124 struct ll_inode_info {
126 cfs_semaphore_t lli_size_sem; /* protect open and change size */
127 void *lli_size_sem_owner;
128 cfs_semaphore_t lli_write_sem;
129 cfs_rw_semaphore_t lli_trunc_sem;
130 char *lli_symlink_name;
133 unsigned long lli_flags;
134 cfs_time_t lli_contention_time;
136 /* this lock protects posix_acl, pending_write_llaps, mmap_cnt */
137 cfs_spinlock_t lli_lock;
138 cfs_list_t lli_close_list;
139 /* handle is to be sent to MDS later on done_writing and setattr.
140 * Open handle data are needed for the recovery to reconstruct
141 * the inode state on the MDS. XXX: recovery is not ready yet. */
142 struct obd_client_handle *lli_pending_och;
144 /* for writepage() only to communicate to fsync */
148 struct posix_acl *lli_posix_acl;
150 /* remote permission hash */
151 cfs_hlist_head_t *lli_remote_perms;
152 unsigned long lli_rmtperm_utime;
153 cfs_semaphore_t lli_rmtperm_sem;
155 cfs_list_t lli_dead_list;
157 cfs_semaphore_t lli_och_sem; /* Protects access to och pointers
158 and their usage counters, also
159 atomicity of check-update of
161 /* We need all three because every inode may be opened in different
163 struct obd_client_handle *lli_mds_read_och;
164 __u64 lli_open_fd_read_count;
165 struct obd_client_handle *lli_mds_write_och;
166 __u64 lli_open_fd_write_count;
167 struct obd_client_handle *lli_mds_exec_och;
168 __u64 lli_open_fd_exec_count;
170 struct inode lli_vfs_inode;
172 /* identifying fields for both metadata and data stacks. */
173 struct lu_fid lli_fid;
174 /* Parent fid for accessing default stripe data on parent directory
175 * for allocating OST objects after a mknod() and later open-by-FID. */
176 struct lu_fid lli_pfid;
177 struct lov_stripe_md *lli_smd;
180 /* open count currently used by capability only, indicate whether
181 * capability needs renewal */
182 cfs_atomic_t lli_open_count;
183 struct obd_capa *lli_mds_capa;
184 cfs_list_t lli_oss_capas;
186 /* metadata statahead */
187 /* protect statahead stuff: lli_opendir_pid, lli_opendir_key, lli_sai,
189 cfs_spinlock_t lli_sa_lock;
191 * "opendir_pid" is the token when lookup/revalid -- I am the owner of
194 pid_t lli_opendir_pid;
196 * since parent-child threads can share the same @file struct,
197 * "opendir_key" is the token when dir close for case of parent exit
198 * before child -- it is me should cleanup the dir readahead. */
199 void *lli_opendir_key;
200 struct ll_statahead_info *lli_sai;
202 struct cl_object *lli_clob;
203 /* the most recent timestamps obtained from mds */
204 struct ost_lvb lli_lvb;
206 * serialize normal readdir and statahead-readdir
208 cfs_semaphore_t lli_readdir_sem;
212 * Locking to guarantee consistency of non-atomic updates to long long i_size,
213 * consistency between file size and KMS, and consistency within
214 * ->lli_smd->lsm_oinfo[]'s.
216 * Implemented by ->lli_size_sem and ->lsm_sem, nested in that order.
219 void ll_inode_size_lock(struct inode *inode, int lock_lsm);
220 void ll_inode_size_unlock(struct inode *inode, int unlock_lsm);
222 // FIXME: replace the name of this with LL_I to conform to kernel stuff
223 // static inline struct ll_inode_info *LL_I(struct inode *inode)
224 static inline struct ll_inode_info *ll_i2info(struct inode *inode)
226 return container_of(inode, struct ll_inode_info, lli_vfs_inode);
229 /* default to about 40meg of readahead on a given system. That much tied
230 * up in 512k readahead requests serviced at 40ms each is about 1GB/s. */
231 #define SBI_DEFAULT_READAHEAD_MAX (40UL << (20 - CFS_PAGE_SHIFT))
233 /* default to read-ahead full files smaller than 2MB on the second read */
234 #define SBI_DEFAULT_READAHEAD_WHOLE_MAX (2UL << (20 - CFS_PAGE_SHIFT))
239 RA_STAT_DISTANT_READPAGE,
240 RA_STAT_MISS_IN_WINDOW,
241 RA_STAT_FAILED_GRAB_PAGE,
242 RA_STAT_FAILED_MATCH,
247 RA_STAT_MAX_IN_FLIGHT,
248 RA_STAT_WRONG_GRAB_PAGE,
253 cfs_atomic_t ra_cur_pages;
254 unsigned long ra_max_pages;
255 unsigned long ra_max_pages_per_file;
256 unsigned long ra_max_read_ahead_whole_pages;
259 /* ra_io_arg will be filled in the beginning of ll_readahead with
260 * ras_lock, then the following ll_read_ahead_pages will read RA
261 * pages according to this arg, all the items in this structure are
262 * counted by page index.
265 unsigned long ria_start; /* start offset of read-ahead*/
266 unsigned long ria_end; /* end offset of read-ahead*/
267 /* If stride read pattern is detected, ria_stoff means where
268 * stride read is started. Note: for normal read-ahead, the
269 * value here is meaningless, and also it will not be accessed*/
271 /* ria_length and ria_pages are the length and pages length in the
272 * stride I/O mode. And they will also be used to check whether
273 * it is stride I/O read-ahead in the read-ahead pages*/
274 unsigned long ria_length;
275 unsigned long ria_pages;
278 /* LL_HIST_MAX=32 causes an overflow */
279 #define LL_HIST_MAX 28
280 #define LL_HIST_START 12 /* buckets start at 2^12 = 4k */
281 #define LL_PROCESS_HIST_MAX 10
282 struct per_process_info {
284 struct obd_histogram pp_r_hist;
285 struct obd_histogram pp_w_hist;
288 /* pp_extents[LL_PROCESS_HIST_MAX] will hold the combined process info */
289 struct ll_rw_extents_info {
290 struct per_process_info pp_extents[LL_PROCESS_HIST_MAX + 1];
293 #define LL_OFFSET_HIST_MAX 100
294 struct ll_rw_process_info {
297 loff_t rw_range_start;
299 loff_t rw_last_file_pos;
301 size_t rw_smallest_extent;
302 size_t rw_largest_extent;
303 struct ll_file_data *rw_last_file;
306 enum stats_track_type {
307 STATS_TRACK_ALL = 0, /* track all processes */
308 STATS_TRACK_PID, /* track process with this pid */
309 STATS_TRACK_PPID, /* track processes with this ppid */
310 STATS_TRACK_GID, /* track processes with this gid */
314 /* flags for sbi->ll_flags */
315 #define LL_SBI_NOLCK 0x01 /* DLM locking disabled (directio-only) */
316 #define LL_SBI_CHECKSUM 0x02 /* checksum each page as it's written */
317 #define LL_SBI_FLOCK 0x04
318 #define LL_SBI_USER_XATTR 0x08 /* support user xattr */
319 #define LL_SBI_ACL 0x10 /* support ACL */
320 #define LL_SBI_RMT_CLIENT 0x40 /* remote client */
321 #define LL_SBI_MDS_CAPA 0x80 /* support mds capa */
322 #define LL_SBI_OSS_CAPA 0x100 /* support oss capa */
323 #define LL_SBI_LOCALFLOCK 0x200 /* Local flocks support by kernel */
324 #define LL_SBI_LRU_RESIZE 0x400 /* lru resize support */
325 #define LL_SBI_LAZYSTATFS 0x800 /* lazystatfs mount option */
326 #define LL_SBI_SOM_PREVIEW 0x1000 /* SOM preview mount option */
327 #define LL_SBI_32BIT_API 0x2000 /* generate 32 bit inodes. */
328 #define LL_SBI_64BIT_HASH 0x4000 /* support 64-bits dir hash/offset */
330 /* default value for ll_sb_info->contention_time */
331 #define SBI_DEFAULT_CONTENTION_SECONDS 60
332 /* default value for lockless_truncate_enable */
333 #define SBI_DEFAULT_LOCKLESS_TRUNCATE_ENABLE 1
334 #define RCE_HASHES 32
336 struct rmtacl_ctl_entry {
338 pid_t rce_key; /* hash key */
339 int rce_ops; /* acl operation type */
342 struct rmtacl_ctl_table {
343 cfs_spinlock_t rct_lock;
344 cfs_list_t rct_entries[RCE_HASHES];
351 pid_t ee_key; /* hash key */
352 struct lu_fid ee_fid;
353 int ee_type; /* ACL type for ACCESS or DEFAULT */
354 ext_acl_xattr_header *ee_acl;
358 cfs_spinlock_t et_lock;
359 cfs_list_t et_entries[EE_HASHES];
364 /* this protects pglist and ra_info. It isn't safe to
365 * grab from interrupt contexts */
366 cfs_spinlock_t ll_lock;
367 cfs_spinlock_t ll_pp_extent_lock; /* Lock for pp_extent entries */
368 cfs_spinlock_t ll_process_lock; /* Lock for ll_rw_process_info */
369 struct obd_uuid ll_sb_uuid;
370 struct obd_export *ll_md_exp;
371 struct obd_export *ll_dt_exp;
372 struct proc_dir_entry* ll_proc_root;
373 struct lu_fid ll_root_fid; /* root object fid */
376 cfs_list_t ll_conn_chain; /* per-conn chain of SBs */
377 struct lustre_client_ocd ll_lco;
379 cfs_list_t ll_orphan_dentry_list; /*please don't ask -p*/
380 struct ll_close_queue *ll_lcq;
382 struct lprocfs_stats *ll_stats; /* lprocfs stats counter */
384 unsigned long ll_async_page_max;
385 unsigned long ll_async_page_count;
387 struct lprocfs_stats *ll_ra_stats;
389 struct ll_ra_info ll_ra_info;
390 unsigned int ll_namelen;
391 struct file_operations *ll_fop;
393 /* =0 - hold lock over whole read/write
394 * >0 - max. chunk to be read/written w/o lock re-acquiring */
395 unsigned long ll_max_rw_chunk;
397 struct lu_site *ll_site;
398 struct cl_device *ll_cl;
400 struct ll_rw_extents_info ll_rw_extents_info;
401 int ll_extent_process_count;
402 struct ll_rw_process_info ll_rw_process_info[LL_PROCESS_HIST_MAX];
403 unsigned int ll_offset_process_count;
404 struct ll_rw_process_info ll_rw_offset_info[LL_OFFSET_HIST_MAX];
405 unsigned int ll_rw_offset_entry_count;
406 enum stats_track_type ll_stats_track_type;
407 int ll_stats_track_id;
410 /* metadata stat-ahead */
411 unsigned int ll_sa_max; /* max statahead RPCs */
412 atomic_t ll_sa_total; /* statahead thread started
414 atomic_t ll_sa_wrong; /* statahead thread stopped for
417 dev_t ll_sdev_orig; /* save s_dev before assign for
419 struct rmtacl_ctl_table ll_rct;
420 struct eacl_table ll_et;
421 struct vfsmount *ll_mnt;
424 #define LL_DEFAULT_MAX_RW_CHUNK (32 * 1024 * 1024)
429 struct task_struct *lrr_reader;
430 cfs_list_t lrr_linkage;
434 * per file-descriptor read-ahead data.
436 struct ll_readahead_state {
437 cfs_spinlock_t ras_lock;
439 * index of the last page that read(2) needed and that wasn't in the
440 * cache. Used by ras_update() to detect seeks.
442 * XXX nikita: if access seeks into cached region, Lustre doesn't see
445 unsigned long ras_last_readpage;
447 * number of pages read after last read-ahead window reset. As window
448 * is reset on each seek, this is effectively a number of consecutive
449 * accesses. Maybe ->ras_accessed_in_window is better name.
451 * XXX nikita: window is also reset (by ras_update()) when Lustre
452 * believes that memory pressure evicts read-ahead pages. In that
453 * case, it probably doesn't make sense to expand window to
454 * PTLRPC_MAX_BRW_PAGES on the third access.
456 unsigned long ras_consecutive_pages;
458 * number of read requests after the last read-ahead window reset
459 * As window is reset on each seek, this is effectively the number
460 * on consecutive read request and is used to trigger read-ahead.
462 unsigned long ras_consecutive_requests;
464 * Parameters of current read-ahead window. Handled by
465 * ras_update(). On the initial access to the file or after a seek,
466 * window is reset to 0. After 3 consecutive accesses, window is
467 * expanded to PTLRPC_MAX_BRW_PAGES. Afterwards, window is enlarged by
468 * PTLRPC_MAX_BRW_PAGES chunks up to ->ra_max_pages.
470 unsigned long ras_window_start, ras_window_len;
472 * Where next read-ahead should start at. This lies within read-ahead
473 * window. Read-ahead window is read in pieces rather than at once
474 * because: 1. lustre limits total number of pages under read-ahead by
475 * ->ra_max_pages (see ll_ra_count_get()), 2. client cannot read pages
476 * not covered by DLM lock.
478 unsigned long ras_next_readahead;
480 * Total number of ll_file_read requests issued, reads originating
481 * due to mmap are not counted in this total. This value is used to
482 * trigger full file read-ahead after multiple reads to a small file.
484 unsigned long ras_requests;
486 * Page index with respect to the current request, these value
487 * will not be accurate when dealing with reads issued via mmap.
489 unsigned long ras_request_index;
491 * list of struct ll_ra_read's one per read(2) call current in
492 * progress against this file descriptor. Used by read-ahead code,
493 * protected by ->ras_lock.
495 cfs_list_t ras_read_beads;
497 * The following 3 items are used for detecting the stride I/O
499 * In stride I/O mode,
500 * ...............|-----data-----|****gap*****|--------|******|....
501 * offset |-stride_pages-|-stride_gap-|
502 * ras_stride_offset = offset;
503 * ras_stride_length = stride_pages + stride_gap;
504 * ras_stride_pages = stride_pages;
505 * Note: all these three items are counted by pages.
507 unsigned long ras_stride_length;
508 unsigned long ras_stride_pages;
509 pgoff_t ras_stride_offset;
511 * number of consecutive stride request count, and it is similar as
512 * ras_consecutive_requests, but used for stride I/O mode.
513 * Note: only more than 2 consecutive stride request are detected,
514 * stride read-ahead will be enable
516 unsigned long ras_consecutive_stride_requests;
524 extern cfs_mem_cache_t *ll_file_data_slab;
525 struct lustre_handle;
526 struct ll_file_data {
527 struct ll_readahead_state fd_ras;
529 struct ccc_grouplock fd_grouplock;
530 struct ll_file_dir fd_dir;
532 struct file *fd_file;
535 struct lov_stripe_md;
537 extern cfs_spinlock_t inode_lock;
539 extern struct proc_dir_entry *proc_lustre_fs_root;
541 static inline struct inode *ll_info2i(struct ll_inode_info *lli)
543 return &lli->lli_vfs_inode;
547 struct inode *icbd_parent;
548 struct dentry **icbd_childp;
552 __u32 ll_i2suppgid(struct inode *i);
553 void ll_i2gids(__u32 *suppgids, struct inode *i1,struct inode *i2);
555 static inline int ll_need_32bit_api(struct ll_sb_info *sbi)
557 #if BITS_PER_LONG == 32
560 return unlikely(cfs_curproc_is_32bit() || (sbi->ll_flags & LL_SBI_32BIT_API));
564 #define LLAP_MAGIC 98764321
566 extern cfs_mem_cache_t *ll_async_page_slab;
567 extern size_t ll_async_page_slab_size;
569 void ll_ra_read_in(struct file *f, struct ll_ra_read *rar);
570 void ll_ra_read_ex(struct file *f, struct ll_ra_read *rar);
571 struct ll_ra_read *ll_ra_read_get(struct file *f);
573 /* llite/lproc_llite.c */
575 int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
576 struct super_block *sb, char *osc, char *mdc);
577 void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi);
578 void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count);
579 void lprocfs_llite_init_vars(struct lprocfs_static_vars *lvars);
581 static inline int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
582 struct super_block *sb, char *osc, char *mdc){return 0;}
583 static inline void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi) {}
584 static void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count) {}
585 static void lprocfs_llite_init_vars(struct lprocfs_static_vars *lvars)
587 memset(lvars, 0, sizeof(*lvars));
593 static inline void ll_put_page(struct page *page)
596 page_cache_release(page);
599 extern struct file_operations ll_dir_operations;
600 extern struct inode_operations ll_dir_inode_operations;
601 struct page *ll_get_dir_page(struct file *filp, struct inode *dir, __u64 hash,
602 int exact, struct ll_dir_chain *chain);
603 int ll_readdir(struct file *filp, void *cookie, filldir_t filldir);
605 int ll_get_mdt_idx(struct inode *inode);
607 int ll_objects_destroy(struct ptlrpc_request *request,
609 struct inode *ll_iget(struct super_block *sb, ino_t hash,
610 struct lustre_md *lic);
611 int ll_md_blocking_ast(struct ldlm_lock *, struct ldlm_lock_desc *,
612 void *data, int flag);
613 struct lookup_intent *ll_convert_intent(struct open_intent *oit,
615 int ll_lookup_it_finish(struct ptlrpc_request *request,
616 struct lookup_intent *it, void *data);
619 int ll_prepare_write(struct file *, struct page *, unsigned from, unsigned to);
620 int ll_commit_write(struct file *, struct page *, unsigned from, unsigned to);
621 int ll_writepage(struct page *page, struct writeback_control *wbc);
622 void ll_removepage(struct page *page);
623 int ll_readpage(struct file *file, struct page *page);
624 void ll_readahead_init(struct inode *inode, struct ll_readahead_state *ras);
625 void ll_truncate(struct inode *inode);
626 int ll_file_punch(struct inode *, loff_t, int);
627 ssize_t ll_file_lockless_io(struct file *, char *, size_t, loff_t *, int);
628 void ll_clear_file_contended(struct inode*);
629 int ll_sync_page_range(struct inode *, struct address_space *, loff_t, size_t);
630 int ll_readahead(const struct lu_env *env, struct cl_io *io, struct ll_readahead_state *ras,
631 struct address_space *mapping, struct cl_page_list *queue, int flags);
634 extern struct file_operations ll_file_operations;
635 extern struct file_operations ll_file_operations_flock;
636 extern struct file_operations ll_file_operations_noflock;
637 extern struct inode_operations ll_file_inode_operations;
638 extern int ll_inode_revalidate_it(struct dentry *, struct lookup_intent *,
640 extern int ll_have_md_lock(struct inode *inode, __u64 bits, ldlm_mode_t l_req_mode);
641 extern ldlm_mode_t ll_take_md_lock(struct inode *inode, __u64 bits,
642 struct lustre_handle *lockh);
643 int __ll_inode_revalidate_it(struct dentry *, struct lookup_intent *, __u64 bits);
644 int ll_revalidate_nd(struct dentry *dentry, struct nameidata *nd);
645 int ll_file_open(struct inode *inode, struct file *file);
646 int ll_file_release(struct inode *inode, struct file *file);
647 int ll_glimpse_ioctl(struct ll_sb_info *sbi,
648 struct lov_stripe_md *lsm, lstat_t *st);
649 void ll_ioepoch_open(struct ll_inode_info *lli, __u64 ioepoch);
650 int ll_local_open(struct file *file,
651 struct lookup_intent *it, struct ll_file_data *fd,
652 struct obd_client_handle *och);
653 int ll_release_openhandle(struct dentry *, struct lookup_intent *);
654 int ll_md_close(struct obd_export *md_exp, struct inode *inode,
656 int ll_md_real_close(struct inode *inode, int flags);
657 void ll_ioepoch_close(struct inode *inode, struct md_op_data *op_data,
658 struct obd_client_handle **och, unsigned long flags);
659 void ll_done_writing_attr(struct inode *inode, struct md_op_data *op_data);
660 int ll_som_update(struct inode *inode, struct md_op_data *op_data);
661 int ll_inode_getattr(struct inode *inode, struct obdo *obdo,
662 __u64 ioepoch, int sync);
663 int ll_md_setattr(struct inode *inode, struct md_op_data *op_data,
664 struct md_open_data **mod);
665 void ll_pack_inode2opdata(struct inode *inode, struct md_op_data *op_data,
666 struct lustre_handle *fh);
667 extern void ll_rw_stats_tally(struct ll_sb_info *sbi, pid_t pid,
668 struct ll_file_data *file, loff_t pos,
669 size_t count, int rw);
670 int ll_getattr_it(struct vfsmount *mnt, struct dentry *de,
671 struct lookup_intent *it, struct kstat *stat);
672 int ll_getattr(struct vfsmount *mnt, struct dentry *de, struct kstat *stat);
673 struct ll_file_data *ll_file_data_get(void);
674 #ifndef HAVE_INODE_PERMISION_2ARGS
675 int ll_inode_permission(struct inode *inode, int mask, struct nameidata *nd);
677 int ll_inode_permission(struct inode *inode, int mask);
679 int ll_lov_setstripe_ea_info(struct inode *inode, struct file *file,
680 int flags, struct lov_user_md *lum,
682 int ll_lov_getstripe_ea_info(struct inode *inode, const char *filename,
683 struct lov_mds_md **lmm, int *lmm_size,
684 struct ptlrpc_request **request);
685 int ll_dir_setstripe(struct inode *inode, struct lov_user_md *lump,
687 int ll_dir_getstripe(struct inode *inode, struct lov_mds_md **lmm,
688 int *lmm_size, struct ptlrpc_request **request);
689 int ll_fsync(struct file *file, struct dentry *dentry, int data);
690 int ll_do_fiemap(struct inode *inode, struct ll_user_fiemap *fiemap,
692 int ll_merge_lvb(struct inode *inode);
693 int ll_get_grouplock(struct inode *inode, struct file *file, unsigned long arg);
694 int ll_put_grouplock(struct inode *inode, struct file *file, unsigned long arg);
695 int ll_fid2path(struct obd_export *exp, void *arg);
700 * protect race ll_find_aliases vs ll_revalidate_it vs ll_unhash_aliases
702 int ll_dops_init(struct dentry *de, int block, int init_sa);
703 extern cfs_spinlock_t ll_lookup_lock;
704 extern struct dentry_operations ll_d_ops;
705 void ll_intent_drop_lock(struct lookup_intent *);
706 void ll_intent_release(struct lookup_intent *);
707 int ll_drop_dentry(struct dentry *dentry);
708 int ll_drop_dentry(struct dentry *dentry);
709 void ll_unhash_aliases(struct inode *);
710 void ll_frob_intent(struct lookup_intent **itp, struct lookup_intent *deft);
711 void ll_lookup_finish_locks(struct lookup_intent *it, struct dentry *dentry);
712 int ll_dcompare(struct dentry *parent, struct qstr *d_name, struct qstr *name);
713 int ll_revalidate_it_finish(struct ptlrpc_request *request,
714 struct lookup_intent *it, struct dentry *de);
716 /* llite/llite_lib.c */
717 extern struct super_operations lustre_super_operations;
719 char *ll_read_opt(const char *opt, char *data);
720 void ll_lli_init(struct ll_inode_info *lli);
721 int ll_fill_super(struct super_block *sb, struct vfsmount *mnt);
722 void ll_put_super(struct super_block *sb);
723 void ll_kill_super(struct super_block *sb);
724 struct inode *ll_inode_from_lock(struct ldlm_lock *lock);
725 void ll_clear_inode(struct inode *inode);
726 int ll_setattr_raw(struct inode *inode, struct iattr *attr);
727 int ll_setattr(struct dentry *de, struct iattr *attr);
728 #ifndef HAVE_STATFS_DENTRY_PARAM
729 int ll_statfs(struct super_block *sb, struct kstatfs *sfs);
731 int ll_statfs(struct dentry *de, struct kstatfs *sfs);
733 int ll_statfs_internal(struct super_block *sb, struct obd_statfs *osfs,
734 __u64 max_age, __u32 flags);
735 void ll_update_inode(struct inode *inode, struct lustre_md *md);
736 void ll_read_inode2(struct inode *inode, void *opaque);
737 void ll_delete_inode(struct inode *inode);
738 int ll_iocontrol(struct inode *inode, struct file *file,
739 unsigned int cmd, unsigned long arg);
740 int ll_flush_ctx(struct inode *inode);
741 #ifdef HAVE_UMOUNTBEGIN_VFSMOUNT
742 void ll_umount_begin(struct vfsmount *vfsmnt, int flags);
744 void ll_umount_begin(struct super_block *sb);
746 int ll_remount_fs(struct super_block *sb, int *flags, char *data);
747 int ll_show_options(struct seq_file *seq, struct vfsmount *vfs);
748 int ll_prep_inode(struct inode **inode, struct ptlrpc_request *req,
749 struct super_block *);
750 void lustre_dump_dentry(struct dentry *, int recur);
751 void lustre_dump_inode(struct inode *);
752 int ll_obd_statfs(struct inode *inode, void *arg);
753 int ll_get_max_mdsize(struct ll_sb_info *sbi, int *max_mdsize);
754 int ll_process_config(struct lustre_cfg *lcfg);
755 struct md_op_data *ll_prep_md_op_data(struct md_op_data *op_data,
756 struct inode *i1, struct inode *i2,
757 const char *name, int namelen,
758 int mode, __u32 opc, void *data);
759 void ll_finish_md_op_data(struct md_op_data *op_data);
761 /* llite/llite_nfs.c */
762 extern struct export_operations lustre_export_operations;
763 __u32 get_uuid2int(const char *name, int len);
765 /* llite/special.c */
766 extern struct inode_operations ll_special_inode_operations;
767 extern struct file_operations ll_special_chr_inode_fops;
768 extern struct file_operations ll_special_chr_file_fops;
769 extern struct file_operations ll_special_blk_inode_fops;
770 extern struct file_operations ll_special_fifo_inode_fops;
771 extern struct file_operations ll_special_fifo_file_fops;
772 extern struct file_operations ll_special_sock_inode_fops;
774 /* llite/symlink.c */
775 extern struct inode_operations ll_fast_symlink_inode_operations;
777 /* llite/llite_close.c */
778 struct ll_close_queue {
779 cfs_spinlock_t lcq_lock;
781 cfs_waitq_t lcq_waitq;
782 cfs_completion_t lcq_comp;
783 cfs_atomic_t lcq_stop;
786 struct ccc_object *cl_inode2ccc(struct inode *inode);
789 void vvp_write_pending (struct ccc_object *club, struct ccc_page *page);
790 void vvp_write_complete(struct ccc_object *club, struct ccc_page *page);
792 /* specific achitecture can implement only part of this list */
793 enum vvp_io_subtype {
796 /** io called from .sendfile */
798 /** io started from splice_{read|write} */
805 enum vvp_io_subtype cui_io_subtype;
809 read_actor_t cui_actor;
813 struct pipe_inode_info *cui_pipe;
814 unsigned int cui_flags;
816 struct vvp_fault_io {
818 * Inode modification time that is checked across DLM
822 struct vm_area_struct *ft_vma;
824 * locked page returned from vvp_io
826 cfs_page_t *ft_vmpage;
827 #ifndef HAVE_VM_OP_FAULT
828 struct vm_nopage_api {
830 * Virtual address at which fault occurred.
832 unsigned long ft_address;
834 * Fault type, as to be supplied to
840 struct vm_fault_api {
844 struct vm_fault *ft_vmf;
846 * fault API used bitflags for return code.
848 unsigned int ft_flags;
854 * Read-ahead state used by read and page-fault IO contexts.
856 struct ll_ra_read cui_bead;
858 * Set when cui_bead has been initialized.
860 int cui_ra_window_set;
862 * Partially truncated page, that vvp_io_trunc_start() keeps locked
865 struct cl_page *cui_partpage;
869 * IO arguments for various VFS I/O interfaces.
872 /** normal/sendfile/splice */
873 enum vvp_io_subtype via_io_subtype;
877 #ifndef HAVE_FILE_WRITEV
878 struct kiocb *via_iocb;
880 struct iovec *via_iov;
881 unsigned long via_nrsegs;
884 read_actor_t via_actor;
888 struct pipe_inode_info *via_pipe;
889 unsigned int via_flags;
894 struct ll_cl_context {
896 struct cl_io *lcc_io;
897 struct cl_page *lcc_page;
898 struct lu_env *lcc_env;
903 struct vvp_thread_info {
904 struct ost_lvb vti_lvb;
905 struct cl_2queue vti_queue;
906 struct iovec vti_local_iov;
907 struct vvp_io_args vti_args;
908 struct ra_io_arg vti_ria;
909 struct kiocb vti_kiocb;
910 struct ll_cl_context vti_io_ctx;
913 static inline struct vvp_thread_info *vvp_env_info(const struct lu_env *env)
915 extern struct lu_context_key vvp_key;
916 struct vvp_thread_info *info;
918 info = lu_context_key_get(&env->le_ctx, &vvp_key);
919 LASSERT(info != NULL);
923 static inline struct vvp_io_args *vvp_env_args(const struct lu_env *env,
924 enum vvp_io_subtype type)
926 struct vvp_io_args *ret = &vvp_env_info(env)->vti_args;
928 ret->via_io_subtype = type;
934 struct vvp_io vs_ios;
937 static inline struct vvp_session *vvp_env_session(const struct lu_env *env)
939 extern struct lu_context_key vvp_session_key;
940 struct vvp_session *ses;
942 ses = lu_context_key_get(env->le_ses, &vvp_session_key);
943 LASSERT(ses != NULL);
947 static inline struct vvp_io *vvp_env_io(const struct lu_env *env)
949 return &vvp_env_session(env)->vs_ios;
952 void ll_queue_done_writing(struct inode *inode, unsigned long flags);
953 void ll_close_thread_shutdown(struct ll_close_queue *lcq);
954 int ll_close_thread_start(struct ll_close_queue **lcq_ret);
956 /* llite/llite_mmap.c */
957 typedef struct rb_root rb_root_t;
958 typedef struct rb_node rb_node_t;
960 struct ll_lock_tree_node;
961 struct ll_lock_tree {
963 cfs_list_t lt_locked_list;
964 struct ll_file_data *lt_fd;
967 int ll_teardown_mmaps(struct address_space *mapping, __u64 first, __u64 last);
968 int ll_file_mmap(struct file * file, struct vm_area_struct * vma);
969 struct ll_lock_tree_node * ll_node_from_inode(struct inode *inode, __u64 start,
970 __u64 end, ldlm_mode_t mode);
971 void policy_from_vma(ldlm_policy_data_t *policy,
972 struct vm_area_struct *vma, unsigned long addr, size_t count);
973 struct vm_area_struct *our_vma(unsigned long addr, size_t count);
975 #define ll_s2sbi(sb) (s2lsi(sb)->lsi_llsbi)
977 /* don't need an addref as the sb_info should be holding one */
978 static inline struct obd_export *ll_s2dtexp(struct super_block *sb)
980 return ll_s2sbi(sb)->ll_dt_exp;
983 /* don't need an addref as the sb_info should be holding one */
984 static inline struct obd_export *ll_s2mdexp(struct super_block *sb)
986 return ll_s2sbi(sb)->ll_md_exp;
989 static inline struct client_obd *sbi2mdc(struct ll_sb_info *sbi)
991 struct obd_device *obd = sbi->ll_md_exp->exp_obd;
997 // FIXME: replace the name of this with LL_SB to conform to kernel stuff
998 static inline struct ll_sb_info *ll_i2sbi(struct inode *inode)
1000 return ll_s2sbi(inode->i_sb);
1003 static inline struct obd_export *ll_i2dtexp(struct inode *inode)
1005 return ll_s2dtexp(inode->i_sb);
1008 static inline struct obd_export *ll_i2mdexp(struct inode *inode)
1010 return ll_s2mdexp(inode->i_sb);
1013 static inline struct lu_fid *ll_inode2fid(struct inode *inode)
1016 LASSERT(inode != NULL);
1017 fid = &ll_i2info(inode)->lli_fid;
1018 LASSERT(fid_is_igif(fid) || fid_ver(fid) == 0);
1022 static inline int ll_mds_max_easize(struct super_block *sb)
1024 return sbi2mdc(ll_s2sbi(sb))->cl_max_mds_easize;
1027 static inline __u64 ll_file_maxbytes(struct inode *inode)
1029 return ll_i2info(inode)->lli_maxbytes;
1033 int ll_setxattr(struct dentry *dentry, const char *name,
1034 const void *value, size_t size, int flags);
1035 ssize_t ll_getxattr(struct dentry *dentry, const char *name,
1036 void *buffer, size_t size);
1037 ssize_t ll_listxattr(struct dentry *dentry, char *buffer, size_t size);
1038 int ll_removexattr(struct dentry *dentry, const char *name);
1040 /* llite/remote_perm.c */
1041 extern cfs_mem_cache_t *ll_remote_perm_cachep;
1042 extern cfs_mem_cache_t *ll_rmtperm_hash_cachep;
1044 cfs_hlist_head_t *alloc_rmtperm_hash(void);
1045 void free_rmtperm_hash(cfs_hlist_head_t *hash);
1046 int ll_update_remote_perm(struct inode *inode, struct mdt_remote_perm *perm);
1047 int lustre_check_remote_perm(struct inode *inode, int mask);
1049 /* llite/llite_capa.c */
1050 extern cfs_timer_t ll_capa_timer;
1052 int ll_capa_thread_start(void);
1053 void ll_capa_thread_stop(void);
1054 void ll_capa_timer_callback(unsigned long unused);
1056 struct obd_capa *ll_add_capa(struct inode *inode, struct obd_capa *ocapa);
1057 int ll_update_capa(struct obd_capa *ocapa, struct lustre_capa *capa);
1059 void ll_capa_open(struct inode *inode);
1060 void ll_capa_close(struct inode *inode);
1062 struct obd_capa *ll_mdscapa_get(struct inode *inode);
1063 struct obd_capa *ll_osscapa_get(struct inode *inode, __u64 opc);
1065 void ll_truncate_free_capa(struct obd_capa *ocapa);
1066 void ll_clear_inode_capas(struct inode *inode);
1067 void ll_print_capa_stat(struct ll_sb_info *sbi);
1069 /* llite/llite_cl.c */
1070 extern struct lu_device_type vvp_device_type;
1073 * Common IO arguments for various VFS I/O interfaces.
1076 int cl_sb_init(struct super_block *sb);
1077 int cl_sb_fini(struct super_block *sb);
1078 int cl_inode_init(struct inode *inode, struct lustre_md *md);
1079 void cl_inode_fini(struct inode *inode);
1081 enum cl_lock_mode vvp_mode_from_vma(struct vm_area_struct *vma);
1082 void ll_io_init(struct cl_io *io, const struct file *file, int write);
1084 void ras_update(struct ll_sb_info *sbi, struct inode *inode,
1085 struct ll_readahead_state *ras, unsigned long index,
1087 void ll_ra_count_put(struct ll_sb_info *sbi, unsigned long len);
1088 int ll_is_file_contended(struct file *file);
1089 void ll_ra_stats_inc(struct address_space *mapping, enum ra_stat which);
1091 /* llite/llite_rmtacl.c */
1092 #ifdef CONFIG_FS_POSIX_ACL
1093 obd_valid rce_ops2valid(int ops);
1094 struct rmtacl_ctl_entry *rct_search(struct rmtacl_ctl_table *rct, pid_t key);
1095 int rct_add(struct rmtacl_ctl_table *rct, pid_t key, int ops);
1096 int rct_del(struct rmtacl_ctl_table *rct, pid_t key);
1097 void rct_init(struct rmtacl_ctl_table *rct);
1098 void rct_fini(struct rmtacl_ctl_table *rct);
1100 void ee_free(struct eacl_entry *ee);
1101 int ee_add(struct eacl_table *et, pid_t key, struct lu_fid *fid, int type,
1102 ext_acl_xattr_header *header);
1103 struct eacl_entry *et_search_del(struct eacl_table *et, pid_t key,
1104 struct lu_fid *fid, int type);
1105 void et_search_free(struct eacl_table *et, pid_t key);
1106 void et_init(struct eacl_table *et);
1107 void et_fini(struct eacl_table *et);
1112 #define LL_SA_RPC_MIN 2
1113 #define LL_SA_RPC_DEF 32
1114 #define LL_SA_RPC_MAX 8192
1116 /* per inode struct, for dir only */
1117 struct ll_statahead_info {
1118 struct inode *sai_inode;
1119 unsigned int sai_generation; /* generation for statahead */
1120 cfs_atomic_t sai_refcount; /* when access this struct, hold
1122 unsigned int sai_sent; /* stat requests sent count */
1123 unsigned int sai_replied; /* stat requests which received
1125 unsigned int sai_max; /* max ahead of lookup */
1126 unsigned int sai_index; /* index of statahead entry */
1127 unsigned int sai_index_next; /* index for the next statahead
1128 * entry to be stated */
1129 unsigned int sai_hit; /* hit count */
1130 unsigned int sai_miss; /* miss count:
1131 * for "ls -al" case, it includes
1132 * hidden dentry miss;
1133 * for "ls -l" case, it does not
1134 * include hidden dentry miss.
1135 * "sai_miss_hidden" is used for
1138 unsigned int sai_consecutive_miss; /* consecutive miss */
1139 unsigned int sai_miss_hidden;/* "ls -al", but first dentry
1140 * is not a hidden one */
1141 unsigned int sai_skip_hidden;/* skipped hidden dentry count */
1142 unsigned int sai_ls_all:1; /* "ls -al", do stat-ahead for
1144 cfs_waitq_t sai_waitq; /* stat-ahead wait queue */
1145 struct ptlrpc_thread sai_thread; /* stat-ahead thread */
1146 cfs_list_t sai_entries_sent; /* entries sent out */
1147 cfs_list_t sai_entries_received; /* entries returned */
1148 cfs_list_t sai_entries_stated; /* entries stated */
1149 pid_t sai_pid; /* pid of statahead itself */
1152 int do_statahead_enter(struct inode *dir, struct dentry **dentry, int lookup);
1153 void ll_statahead_exit(struct inode *dir, struct dentry *dentry, int result);
1154 void ll_stop_statahead(struct inode *dir, void *key);
1157 void ll_statahead_mark(struct inode *dir, struct dentry *dentry)
1159 struct ll_inode_info *lli;
1160 struct ll_dentry_data *ldd = ll_d2d(dentry);
1162 /* dentry has been move to other directory, no need mark */
1163 if (unlikely(dir != dentry->d_parent->d_inode))
1166 lli = ll_i2info(dir);
1167 /* not the same process, don't mark */
1168 if (lli->lli_opendir_pid != cfs_curproc_pid())
1171 cfs_spin_lock(&lli->lli_sa_lock);
1172 if (likely(lli->lli_sai != NULL && ldd != NULL))
1173 ldd->lld_sa_generation = lli->lli_sai->sai_generation;
1174 cfs_spin_unlock(&lli->lli_sa_lock);
1178 int ll_statahead_enter(struct inode *dir, struct dentry **dentryp, int lookup)
1180 struct ll_inode_info *lli;
1181 struct ll_dentry_data *ldd = ll_d2d(*dentryp);
1183 if (unlikely(dir == NULL))
1186 if (ll_i2sbi(dir)->ll_sa_max == 0)
1189 lli = ll_i2info(dir);
1190 /* not the same process, don't statahead */
1191 if (lli->lli_opendir_pid != cfs_curproc_pid())
1195 * When "ls" a dentry, the system trigger more than once "revalidate" or
1196 * "lookup", for "getattr", for "getxattr", and maybe for others.
1197 * Under patchless client mode, the operation intent is not accurate,
1198 * it maybe misguide the statahead thread. For example:
1199 * The "revalidate" call for "getattr" and "getxattr" of a dentry maybe
1200 * have the same operation intent -- "IT_GETATTR".
1201 * In fact, one dentry should has only one chance to interact with the
1202 * statahead thread, otherwise the statahead windows will be confused.
1203 * The solution is as following:
1204 * Assign "lld_sa_generation" with "sai_generation" when a dentry
1205 * "IT_GETATTR" for the first time, and the subsequent "IT_GETATTR"
1206 * will bypass interacting with statahead thread for checking:
1207 * "lld_sa_generation == lli_sai->sai_generation"
1209 if (ldd && lli->lli_sai &&
1210 ldd->lld_sa_generation == lli->lli_sai->sai_generation)
1213 return do_statahead_enter(dir, dentryp, lookup);
1216 /* llite ioctl register support rountine */
1223 #define LLIOC_MAX_CMD 256
1226 * Rules to write a callback function:
1229 * @magic: Dynamic ioctl call routine will feed this vaule with the pointer
1230 * returned to ll_iocontrol_register. Callback functions should use this
1231 * data to check the potential collasion of ioctl cmd. If collasion is
1232 * found, callback function should return LLIOC_CONT.
1233 * @rcp: The result of ioctl command.
1236 * If @magic matches the pointer returned by ll_iocontrol_data, the
1237 * callback should return LLIOC_STOP; return LLIOC_STOP otherwise.
1239 typedef enum llioc_iter (*llioc_callback_t)(struct inode *inode,
1240 struct file *file, unsigned int cmd, unsigned long arg,
1241 void *magic, int *rcp);
1243 enum llioc_iter ll_iocontrol_call(struct inode *inode, struct file *file,
1244 unsigned int cmd, unsigned long arg, int *rcp);
1246 /* export functions */
1247 /* Register ioctl block dynamatically for a regular file.
1249 * @cmd: the array of ioctl command set
1250 * @count: number of commands in the @cmd
1251 * @cb: callback function, it will be called if an ioctl command is found to
1252 * belong to the command list @cmd.
1255 * A magic pointer will be returned if success;
1256 * otherwise, NULL will be returned.
1258 void *ll_iocontrol_register(llioc_callback_t cb, int count, unsigned int *cmd);
1259 void ll_iocontrol_unregister(void *magic);
1263 /* lclient compat stuff */
1264 #define cl_inode_info ll_inode_info
1265 #define cl_i2info(info) ll_i2info(info)
1266 #define cl_inode_mode(inode) ((inode)->i_mode)
1267 #define cl_i2sbi ll_i2sbi
1269 static inline void cl_isize_lock(struct inode *inode, int lsmlock)
1271 ll_inode_size_lock(inode, lsmlock);
1274 static inline void cl_isize_unlock(struct inode *inode, int lsmlock)
1276 ll_inode_size_unlock(inode, lsmlock);
1279 static inline void cl_isize_write_nolock(struct inode *inode, loff_t kms)
1281 LASSERT_SEM_LOCKED(&ll_i2info(inode)->lli_size_sem);
1282 i_size_write(inode, kms);
1285 static inline void cl_isize_write(struct inode *inode, loff_t kms)
1287 ll_inode_size_lock(inode, 0);
1288 i_size_write(inode, kms);
1289 ll_inode_size_unlock(inode, 0);
1292 #define cl_isize_read(inode) i_size_read(inode)
1294 static inline int cl_merge_lvb(struct inode *inode)
1296 return ll_merge_lvb(inode);
1299 #define cl_inode_atime(inode) LTIME_S((inode)->i_atime)
1300 #define cl_inode_ctime(inode) LTIME_S((inode)->i_ctime)
1301 #define cl_inode_mtime(inode) LTIME_S((inode)->i_mtime)
1303 struct obd_capa *cl_capa_lookup(struct inode *inode, enum cl_req_type crt);
1305 /** direct write pages */
1306 struct ll_dio_pages {
1307 /** page array to be written. we don't support
1308 * partial pages except the last one. */
1309 struct page **ldp_pages;
1310 /* offset of each page */
1311 loff_t *ldp_offsets;
1312 /** if ldp_offsets is NULL, it means a sequential
1313 * pages to be written, then this is the file offset
1314 * of the * first page. */
1315 loff_t ldp_start_offset;
1316 /** how many bytes are to be written. */
1318 /** # of pages in the array. */
1322 extern ssize_t ll_direct_rw_pages(const struct lu_env *env, struct cl_io *io,
1323 int rw, struct inode *inode,
1324 struct ll_dio_pages *pv);
1326 static inline int ll_file_nolock(const struct file *file)
1328 struct ll_file_data *fd = LUSTRE_FPRIVATE(file);
1329 struct inode *inode = file->f_dentry->d_inode;
1331 LASSERT(fd != NULL);
1332 return ((fd->fd_flags & LL_FILE_IGNORE_LOCK) ||
1333 (ll_i2sbi(inode)->ll_flags & LL_SBI_NOLCK));
1335 #endif /* LLITE_INTERNAL_H */