4 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 only,
8 * as published by the Free Software Foundation.
10 * This program is distributed in the hope that it will be useful, but
11 * WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 * General Public License version 2 for more details (a copy is included
14 * in the LICENSE file that accompanied this code).
16 * You should have received a copy of the GNU General Public License
17 * version 2 along with this program; If not, see
18 * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
20 * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
21 * CA 95054 USA or visit www.sun.com if you need additional information or
27 * Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved.
28 * Use is subject to license terms.
30 * Copyright (c) 2011, 2012, Whamcloud, Inc.
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>
50 #include <lustre_mdc.h>
51 #include <linux/lustre_intent.h>
57 #ifndef VM_FAULT_RETRY
58 #define VM_FAULT_RETRY 0
61 /* Kernel 3.1 kills LOOKUP_CONTINUE, LOOKUP_PARENT is equivalent to it.
62 * seem kernel commit 49084c3bb2055c401f3493c13edae14d49128ca0 */
63 #ifndef LOOKUP_CONTINUE
64 #define LOOKUP_CONTINUE LOOKUP_PARENT
67 /** Only used on client-side for indicating the tail of dir hash/offset. */
68 #define LL_DIR_END_OFF 0x7fffffffffffffffULL
69 #define LL_DIR_END_OFF_32BIT 0x7fffffffUL
71 #ifndef DCACHE_LUSTRE_INVALID
72 #define DCACHE_LUSTRE_INVALID 0x4000000
75 #define LL_IT2STR(it) ((it) ? ldlm_it2str((it)->it_op) : "0")
76 #define LUSTRE_FPRIVATE(file) ((file)->private_data)
78 struct ll_dentry_data {
81 struct obd_client_handle lld_cwd_och;
82 struct obd_client_handle lld_mnt_och;
83 struct lookup_intent *lld_it;
84 unsigned int lld_sa_generation;
87 #define ll_d2d(de) ((struct ll_dentry_data*)((de)->d_fsdata))
89 extern struct file_operations ll_pgcache_seq_fops;
91 #define LLI_INODE_MAGIC 0x111d0de5
92 #define LLI_INODE_DEAD 0xdeadd00d
94 /* remote client permission cache */
95 #define REMOTE_PERM_HASHSIZE 16
97 struct ll_getname_data {
98 char *lgd_name; /* points to a buffer with NAME_MAX+1 size */
99 struct lu_fid lgd_fid; /* target fid we are looking for */
100 int lgd_found; /* inode matched? */
103 /* llite setxid/access permission for user on remote client */
104 struct ll_remote_perm {
105 cfs_hlist_node_t lrp_list;
110 int lrp_access_perm; /* MAY_READ/WRITE/EXEC, this
111 is access permission with
112 lrp_fsuid/lrp_fsgid. */
116 /* MDS has an authority for the Size-on-MDS attributes. */
117 LLIF_MDS_SIZE_LOCK = (1 << 0),
118 /* Epoch close is postponed. */
119 LLIF_EPOCH_PENDING = (1 << 1),
120 /* DONE WRITING is allowed. */
121 LLIF_DONE_WRITING = (1 << 2),
122 /* Sizeon-on-MDS attributes are changed. An attribute update needs to
124 LLIF_SOM_DIRTY = (1 << 3),
125 /* File is contented */
126 LLIF_CONTENDED = (1 << 4),
127 /* Truncate uses server lock for this file */
128 LLIF_SRVLOCK = (1 << 5),
132 struct ll_inode_info {
133 __u32 lli_inode_magic;
137 cfs_spinlock_t lli_lock;
138 struct posix_acl *lli_posix_acl;
140 cfs_hlist_head_t *lli_remote_perms;
141 cfs_mutex_t lli_rmtperm_mutex;
143 /* identifying fields for both metadata and data stacks. */
144 struct lu_fid lli_fid;
145 /* Parent fid for accessing default stripe data on parent directory
146 * for allocating OST objects after a mknod() and later open-by-FID. */
147 struct lu_fid lli_pfid;
149 cfs_list_t lli_close_list;
150 cfs_list_t lli_oss_capas;
151 /* open count currently used by capability only, indicate whether
152 * capability needs renewal */
153 cfs_atomic_t lli_open_count;
154 struct obd_capa *lli_mds_capa;
155 cfs_time_t lli_rmtperm_time;
157 /* handle is to be sent to MDS later on done_writing and setattr.
158 * Open handle data are needed for the recovery to reconstruct
159 * the inode state on the MDS. XXX: recovery is not ready yet. */
160 struct obd_client_handle *lli_pending_och;
162 /* We need all three because every inode may be opened in different
164 struct obd_client_handle *lli_mds_read_och;
165 struct obd_client_handle *lli_mds_write_och;
166 struct obd_client_handle *lli_mds_exec_och;
167 __u64 lli_open_fd_read_count;
168 __u64 lli_open_fd_write_count;
169 __u64 lli_open_fd_exec_count;
170 /* Protects access to och pointers and their usage counters, also
171 * atomicity of check-update of lli_has_smd */
172 cfs_mutex_t lli_och_mutex;
174 struct inode lli_vfs_inode;
176 /* the most recent timestamps obtained from mds */
177 struct ost_lvb lli_lvb;
178 cfs_spinlock_t lli_agl_lock;
180 /* Try to make the d::member and f::member are aligned. Before using
181 * these members, make clear whether it is directory or not. */
185 /* serialize normal readdir and statahead-readdir. */
186 cfs_mutex_t d_readdir_mutex;
188 /* metadata statahead */
189 /* since parent-child threads can share the same @file
190 * struct, "opendir_key" is the token when dir close for
191 * case of parent exit before child -- it is me should
192 * cleanup the dir readahead. */
194 struct ll_statahead_info *d_sai;
196 struct posix_acl *d_def_acl;
197 /* protect statahead stuff. */
198 cfs_spinlock_t d_sa_lock;
199 /* "opendir_pid" is the token when lookup/revalid
200 * -- I am the owner of dir statahead. */
204 #define lli_readdir_mutex u.d.d_readdir_mutex
205 #define lli_opendir_key u.d.d_opendir_key
206 #define lli_sai u.d.d_sai
207 #define lli_sa_pos u.d.d_sa_pos
208 #define lli_def_acl u.d.d_def_acl
209 #define lli_sa_lock u.d.d_sa_lock
210 #define lli_opendir_pid u.d.d_opendir_pid
212 /* for non-directory */
214 cfs_semaphore_t f_size_sem;
215 void *f_size_sem_owner;
216 char *f_symlink_name;
219 * cfs_rw_semaphore_t {
220 * signed long count; // align u.d.d_def_acl
221 * cfs_spinlock_t wait_lock; // align u.d.d_sa_lock
222 * struct list_head wait_list;
225 cfs_rw_semaphore_t f_trunc_sem;
226 cfs_mutex_t f_write_mutex;
228 cfs_rw_semaphore_t f_glimpse_sem;
229 cfs_time_t f_glimpse_time;
230 cfs_list_t f_agl_list;
233 /* for writepage() only to communicate to fsync */
237 * whenever a process try to read/write the file, the
238 * jobid of the process will be saved here, and it'll
239 * be packed into the write PRC when flush later.
241 * so the read/write statistics for jobid will not be
242 * accurate if the file is shared by different jobs.
244 char f_jobid[JOBSTATS_JOBID_SIZE];
247 #define lli_size_sem u.f.f_size_sem
248 #define lli_size_sem_owner u.f.f_size_sem_owner
249 #define lli_symlink_name u.f.f_symlink_name
250 #define lli_maxbytes u.f.f_maxbytes
251 #define lli_trunc_sem u.f.f_trunc_sem
252 #define lli_write_mutex u.f.f_write_mutex
253 #define lli_glimpse_sem u.f.f_glimpse_sem
254 #define lli_glimpse_time u.f.f_glimpse_time
255 #define lli_agl_list u.f.f_agl_list
256 #define lli_agl_index u.f.f_agl_index
257 #define lli_async_rc u.f.f_async_rc
258 #define lli_jobid u.f.f_jobid
262 /* XXX: For following frequent used members, although they maybe special
263 * used for non-directory object, it is some time-wasting to check
264 * whether the object is directory or not before using them. On the
265 * other hand, currently, sizeof(f) > sizeof(d), it cannot reduce
266 * the "ll_inode_info" size even if moving those members into u.f.
267 * So keep them out side.
269 * In the future, if more members are added only for directory,
270 * some of the following members can be moved into u.f.
273 struct cl_object *lli_clob;
275 /* mutex to request for layout lock exclusively. */
276 cfs_mutex_t lli_layout_mutex;
280 * Locking to guarantee consistency of non-atomic updates to long long i_size,
281 * consistency between file size and KMS.
283 * Implemented by ->lli_size_sem and ->lsm_lock, nested in that order.
286 void ll_inode_size_lock(struct inode *inode);
287 void ll_inode_size_unlock(struct inode *inode);
289 // FIXME: replace the name of this with LL_I to conform to kernel stuff
290 // static inline struct ll_inode_info *LL_I(struct inode *inode)
291 static inline struct ll_inode_info *ll_i2info(struct inode *inode)
293 return container_of(inode, struct ll_inode_info, lli_vfs_inode);
296 /* default to about 40meg of readahead on a given system. That much tied
297 * up in 512k readahead requests serviced at 40ms each is about 1GB/s. */
298 #define SBI_DEFAULT_READAHEAD_MAX (40UL << (20 - CFS_PAGE_SHIFT))
300 /* default to read-ahead full files smaller than 2MB on the second read */
301 #define SBI_DEFAULT_READAHEAD_WHOLE_MAX (2UL << (20 - CFS_PAGE_SHIFT))
306 RA_STAT_DISTANT_READPAGE,
307 RA_STAT_MISS_IN_WINDOW,
308 RA_STAT_FAILED_GRAB_PAGE,
309 RA_STAT_FAILED_MATCH,
314 RA_STAT_MAX_IN_FLIGHT,
315 RA_STAT_WRONG_GRAB_PAGE,
320 cfs_atomic_t ra_cur_pages;
321 unsigned long ra_max_pages;
322 unsigned long ra_max_pages_per_file;
323 unsigned long ra_max_read_ahead_whole_pages;
326 /* ra_io_arg will be filled in the beginning of ll_readahead with
327 * ras_lock, then the following ll_read_ahead_pages will read RA
328 * pages according to this arg, all the items in this structure are
329 * counted by page index.
332 unsigned long ria_start; /* start offset of read-ahead*/
333 unsigned long ria_end; /* end offset of read-ahead*/
334 /* If stride read pattern is detected, ria_stoff means where
335 * stride read is started. Note: for normal read-ahead, the
336 * value here is meaningless, and also it will not be accessed*/
338 /* ria_length and ria_pages are the length and pages length in the
339 * stride I/O mode. And they will also be used to check whether
340 * it is stride I/O read-ahead in the read-ahead pages*/
341 unsigned long ria_length;
342 unsigned long ria_pages;
345 /* LL_HIST_MAX=32 causes an overflow */
346 #define LL_HIST_MAX 28
347 #define LL_HIST_START 12 /* buckets start at 2^12 = 4k */
348 #define LL_PROCESS_HIST_MAX 10
349 struct per_process_info {
351 struct obd_histogram pp_r_hist;
352 struct obd_histogram pp_w_hist;
355 /* pp_extents[LL_PROCESS_HIST_MAX] will hold the combined process info */
356 struct ll_rw_extents_info {
357 struct per_process_info pp_extents[LL_PROCESS_HIST_MAX + 1];
360 #define LL_OFFSET_HIST_MAX 100
361 struct ll_rw_process_info {
364 loff_t rw_range_start;
366 loff_t rw_last_file_pos;
368 size_t rw_smallest_extent;
369 size_t rw_largest_extent;
370 struct ll_file_data *rw_last_file;
373 enum stats_track_type {
374 STATS_TRACK_ALL = 0, /* track all processes */
375 STATS_TRACK_PID, /* track process with this pid */
376 STATS_TRACK_PPID, /* track processes with this ppid */
377 STATS_TRACK_GID, /* track processes with this gid */
381 /* flags for sbi->ll_flags */
382 #define LL_SBI_NOLCK 0x01 /* DLM locking disabled (directio-only) */
383 #define LL_SBI_CHECKSUM 0x02 /* checksum each page as it's written */
384 #define LL_SBI_FLOCK 0x04
385 #define LL_SBI_USER_XATTR 0x08 /* support user xattr */
386 #define LL_SBI_ACL 0x10 /* support ACL */
387 #define LL_SBI_RMT_CLIENT 0x40 /* remote client */
388 #define LL_SBI_MDS_CAPA 0x80 /* support mds capa */
389 #define LL_SBI_OSS_CAPA 0x100 /* support oss capa */
390 #define LL_SBI_LOCALFLOCK 0x200 /* Local flocks support by kernel */
391 #define LL_SBI_LRU_RESIZE 0x400 /* lru resize support */
392 #define LL_SBI_LAZYSTATFS 0x800 /* lazystatfs mount option */
393 #define LL_SBI_SOM_PREVIEW 0x1000 /* SOM preview mount option */
394 #define LL_SBI_32BIT_API 0x2000 /* generate 32 bit inodes. */
395 #define LL_SBI_64BIT_HASH 0x4000 /* support 64-bits dir hash/offset */
396 #define LL_SBI_AGL_ENABLED 0x8000 /* enable agl */
397 #define LL_SBI_VERBOSE 0x10000 /* verbose mount/umount */
398 #define LL_SBI_LAYOUT_LOCK 0x20000 /* layout lock support */
400 /* default value for ll_sb_info->contention_time */
401 #define SBI_DEFAULT_CONTENTION_SECONDS 60
402 /* default value for lockless_truncate_enable */
403 #define SBI_DEFAULT_LOCKLESS_TRUNCATE_ENABLE 1
404 #define RCE_HASHES 32
406 struct rmtacl_ctl_entry {
408 pid_t rce_key; /* hash key */
409 int rce_ops; /* acl operation type */
412 struct rmtacl_ctl_table {
413 cfs_spinlock_t rct_lock;
414 cfs_list_t rct_entries[RCE_HASHES];
421 pid_t ee_key; /* hash key */
422 struct lu_fid ee_fid;
423 int ee_type; /* ACL type for ACCESS or DEFAULT */
424 ext_acl_xattr_header *ee_acl;
428 cfs_spinlock_t et_lock;
429 cfs_list_t et_entries[EE_HASHES];
434 /* this protects pglist and ra_info. It isn't safe to
435 * grab from interrupt contexts */
436 cfs_spinlock_t ll_lock;
437 cfs_spinlock_t ll_pp_extent_lock; /* Lock for pp_extent entries */
438 cfs_spinlock_t ll_process_lock; /* Lock for ll_rw_process_info */
439 struct obd_uuid ll_sb_uuid;
440 struct obd_export *ll_md_exp;
441 struct obd_export *ll_dt_exp;
442 struct proc_dir_entry* ll_proc_root;
443 struct lu_fid ll_root_fid; /* root object fid */
446 cfs_list_t ll_conn_chain; /* per-conn chain of SBs */
447 struct lustre_client_ocd ll_lco;
449 cfs_list_t ll_orphan_dentry_list; /*please don't ask -p*/
450 struct ll_close_queue *ll_lcq;
452 struct lprocfs_stats *ll_stats; /* lprocfs stats counter */
454 unsigned long ll_async_page_max;
455 unsigned long ll_async_page_count;
457 struct lprocfs_stats *ll_ra_stats;
459 struct ll_ra_info ll_ra_info;
460 unsigned int ll_namelen;
461 struct file_operations *ll_fop;
463 /* =0 - hold lock over whole read/write
464 * >0 - max. chunk to be read/written w/o lock re-acquiring */
465 unsigned long ll_max_rw_chunk;
466 unsigned int ll_md_brw_size; /* used by readdir */
468 struct lu_site *ll_site;
469 struct cl_device *ll_cl;
471 struct ll_rw_extents_info ll_rw_extents_info;
472 int ll_extent_process_count;
473 struct ll_rw_process_info ll_rw_process_info[LL_PROCESS_HIST_MAX];
474 unsigned int ll_offset_process_count;
475 struct ll_rw_process_info ll_rw_offset_info[LL_OFFSET_HIST_MAX];
476 unsigned int ll_rw_offset_entry_count;
477 int ll_stats_track_id;
478 enum stats_track_type ll_stats_track_type;
481 /* metadata stat-ahead */
482 unsigned int ll_sa_max; /* max statahead RPCs */
483 atomic_t ll_sa_total; /* statahead thread started
485 atomic_t ll_sa_wrong; /* statahead thread stopped for
487 atomic_t ll_agl_total; /* AGL thread started count */
489 dev_t ll_sdev_orig; /* save s_dev before assign for
491 struct rmtacl_ctl_table ll_rct;
492 struct eacl_table ll_et;
493 struct vfsmount *ll_mnt;
496 #define LL_DEFAULT_MAX_RW_CHUNK (32 * 1024 * 1024)
501 struct task_struct *lrr_reader;
502 cfs_list_t lrr_linkage;
506 * per file-descriptor read-ahead data.
508 struct ll_readahead_state {
509 cfs_spinlock_t ras_lock;
511 * index of the last page that read(2) needed and that wasn't in the
512 * cache. Used by ras_update() to detect seeks.
514 * XXX nikita: if access seeks into cached region, Lustre doesn't see
517 unsigned long ras_last_readpage;
519 * number of pages read after last read-ahead window reset. As window
520 * is reset on each seek, this is effectively a number of consecutive
521 * accesses. Maybe ->ras_accessed_in_window is better name.
523 * XXX nikita: window is also reset (by ras_update()) when Lustre
524 * believes that memory pressure evicts read-ahead pages. In that
525 * case, it probably doesn't make sense to expand window to
526 * PTLRPC_MAX_BRW_PAGES on the third access.
528 unsigned long ras_consecutive_pages;
530 * number of read requests after the last read-ahead window reset
531 * As window is reset on each seek, this is effectively the number
532 * on consecutive read request and is used to trigger read-ahead.
534 unsigned long ras_consecutive_requests;
536 * Parameters of current read-ahead window. Handled by
537 * ras_update(). On the initial access to the file or after a seek,
538 * window is reset to 0. After 3 consecutive accesses, window is
539 * expanded to PTLRPC_MAX_BRW_PAGES. Afterwards, window is enlarged by
540 * PTLRPC_MAX_BRW_PAGES chunks up to ->ra_max_pages.
542 unsigned long ras_window_start, ras_window_len;
544 * Where next read-ahead should start at. This lies within read-ahead
545 * window. Read-ahead window is read in pieces rather than at once
546 * because: 1. lustre limits total number of pages under read-ahead by
547 * ->ra_max_pages (see ll_ra_count_get()), 2. client cannot read pages
548 * not covered by DLM lock.
550 unsigned long ras_next_readahead;
552 * Total number of ll_file_read requests issued, reads originating
553 * due to mmap are not counted in this total. This value is used to
554 * trigger full file read-ahead after multiple reads to a small file.
556 unsigned long ras_requests;
558 * Page index with respect to the current request, these value
559 * will not be accurate when dealing with reads issued via mmap.
561 unsigned long ras_request_index;
563 * list of struct ll_ra_read's one per read(2) call current in
564 * progress against this file descriptor. Used by read-ahead code,
565 * protected by ->ras_lock.
567 cfs_list_t ras_read_beads;
569 * The following 3 items are used for detecting the stride I/O
571 * In stride I/O mode,
572 * ...............|-----data-----|****gap*****|--------|******|....
573 * offset |-stride_pages-|-stride_gap-|
574 * ras_stride_offset = offset;
575 * ras_stride_length = stride_pages + stride_gap;
576 * ras_stride_pages = stride_pages;
577 * Note: all these three items are counted by pages.
579 unsigned long ras_stride_length;
580 unsigned long ras_stride_pages;
581 pgoff_t ras_stride_offset;
583 * number of consecutive stride request count, and it is similar as
584 * ras_consecutive_requests, but used for stride I/O mode.
585 * Note: only more than 2 consecutive stride request are detected,
586 * stride read-ahead will be enable
588 unsigned long ras_consecutive_stride_requests;
596 extern cfs_mem_cache_t *ll_file_data_slab;
597 struct lustre_handle;
598 struct ll_file_data {
599 struct ll_readahead_state fd_ras;
601 struct ccc_grouplock fd_grouplock;
602 struct ll_file_dir fd_dir;
604 struct file *fd_file;
605 /* Indicate whether need to report failure when close.
606 * true: failure is known, not report again.
607 * false: unknown failure, should report. */
608 bool fd_write_failed;
611 struct lov_stripe_md;
613 extern cfs_spinlock_t inode_lock;
615 extern struct proc_dir_entry *proc_lustre_fs_root;
617 static inline struct inode *ll_info2i(struct ll_inode_info *lli)
619 return &lli->lli_vfs_inode;
623 struct inode *icbd_parent;
624 struct dentry **icbd_childp;
628 __u32 ll_i2suppgid(struct inode *i);
629 void ll_i2gids(__u32 *suppgids, struct inode *i1,struct inode *i2);
631 static inline int ll_need_32bit_api(struct ll_sb_info *sbi)
633 #if BITS_PER_LONG == 32
636 return unlikely(cfs_curproc_is_32bit() || (sbi->ll_flags & LL_SBI_32BIT_API));
640 #define LLAP_MAGIC 98764321
642 extern cfs_mem_cache_t *ll_async_page_slab;
643 extern size_t ll_async_page_slab_size;
645 void ll_ra_read_in(struct file *f, struct ll_ra_read *rar);
646 void ll_ra_read_ex(struct file *f, struct ll_ra_read *rar);
647 struct ll_ra_read *ll_ra_read_get(struct file *f);
649 /* llite/lproc_llite.c */
651 int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
652 struct super_block *sb, char *osc, char *mdc);
653 void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi);
654 void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count);
655 void lprocfs_llite_init_vars(struct lprocfs_static_vars *lvars);
657 static inline int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
658 struct super_block *sb, char *osc, char *mdc){return 0;}
659 static inline void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi) {}
660 static void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count) {}
661 static void lprocfs_llite_init_vars(struct lprocfs_static_vars *lvars)
663 memset(lvars, 0, sizeof(*lvars));
669 void ll_release_page(struct page *page, int remove);
670 extern struct file_operations ll_dir_operations;
671 extern struct inode_operations ll_dir_inode_operations;
672 struct page *ll_get_dir_page(struct file *filp, struct inode *dir, __u64 hash,
673 struct ll_dir_chain *chain);
674 int ll_readdir(struct file *filp, void *cookie, filldir_t filldir);
676 int ll_get_mdt_idx(struct inode *inode);
677 char *ll_get_fsname(struct inode *inode);
679 int ll_objects_destroy(struct ptlrpc_request *request,
681 struct inode *ll_iget(struct super_block *sb, ino_t hash,
682 struct lustre_md *lic);
683 int ll_md_blocking_ast(struct ldlm_lock *, struct ldlm_lock_desc *,
684 void *data, int flag);
685 struct lookup_intent *ll_convert_intent(struct open_intent *oit,
687 int ll_lookup_it_finish(struct ptlrpc_request *request,
688 struct lookup_intent *it, void *data);
689 struct dentry *ll_splice_alias(struct inode *inode, struct dentry *de);
692 int ll_prepare_write(struct file *, struct page *, unsigned from, unsigned to);
693 int ll_commit_write(struct file *, struct page *, unsigned from, unsigned to);
694 int ll_writepage(struct page *page, struct writeback_control *wbc);
695 int ll_writepages(struct address_space *, struct writeback_control *wbc);
696 void ll_removepage(struct page *page);
697 int ll_readpage(struct file *file, struct page *page);
698 void ll_readahead_init(struct inode *inode, struct ll_readahead_state *ras);
699 void ll_truncate(struct inode *inode);
700 int ll_file_punch(struct inode *, loff_t, int);
701 ssize_t ll_file_lockless_io(struct file *, char *, size_t, loff_t *, int);
702 void ll_clear_file_contended(struct inode*);
703 int ll_sync_page_range(struct inode *, struct address_space *, loff_t, size_t);
704 int ll_readahead(const struct lu_env *env, struct cl_io *io,
705 struct ll_readahead_state *ras, struct address_space *mapping,
706 struct cl_page_list *queue, int flags);
709 extern struct file_operations ll_file_operations;
710 extern struct file_operations ll_file_operations_flock;
711 extern struct file_operations ll_file_operations_noflock;
712 extern struct inode_operations ll_file_inode_operations;
713 extern int ll_inode_revalidate_it(struct dentry *, struct lookup_intent *,
715 extern int ll_have_md_lock(struct inode *inode, __u64 *bits,
716 ldlm_mode_t l_req_mode);
717 extern ldlm_mode_t ll_take_md_lock(struct inode *inode, __u64 bits,
718 struct lustre_handle *lockh);
719 int __ll_inode_revalidate_it(struct dentry *, struct lookup_intent *,
721 int ll_revalidate_nd(struct dentry *dentry, struct nameidata *nd);
722 int ll_file_open(struct inode *inode, struct file *file);
723 int ll_file_release(struct inode *inode, struct file *file);
724 int ll_glimpse_ioctl(struct ll_sb_info *sbi,
725 struct lov_stripe_md *lsm, lstat_t *st);
726 void ll_ioepoch_open(struct ll_inode_info *lli, __u64 ioepoch);
727 int ll_local_open(struct file *file,
728 struct lookup_intent *it, struct ll_file_data *fd,
729 struct obd_client_handle *och);
730 int ll_release_openhandle(struct dentry *, struct lookup_intent *);
731 int ll_md_close(struct obd_export *md_exp, struct inode *inode,
733 int ll_md_real_close(struct inode *inode, int flags);
734 void ll_ioepoch_close(struct inode *inode, struct md_op_data *op_data,
735 struct obd_client_handle **och, unsigned long flags);
736 void ll_done_writing_attr(struct inode *inode, struct md_op_data *op_data);
737 int ll_som_update(struct inode *inode, struct md_op_data *op_data);
738 int ll_inode_getattr(struct inode *inode, struct obdo *obdo,
739 __u64 ioepoch, int sync);
740 int ll_md_setattr(struct dentry *dentry, struct md_op_data *op_data,
741 struct md_open_data **mod);
742 void ll_pack_inode2opdata(struct inode *inode, struct md_op_data *op_data,
743 struct lustre_handle *fh);
744 extern void ll_rw_stats_tally(struct ll_sb_info *sbi, pid_t pid,
745 struct ll_file_data *file, loff_t pos,
746 size_t count, int rw);
747 int ll_getattr_it(struct vfsmount *mnt, struct dentry *de,
748 struct lookup_intent *it, struct kstat *stat);
749 int ll_getattr(struct vfsmount *mnt, struct dentry *de, struct kstat *stat);
750 struct ll_file_data *ll_file_data_get(void);
751 #ifdef HAVE_GENERIC_PERMISSION_4ARGS
752 int ll_inode_permission(struct inode *inode, int mask, unsigned int flags);
754 # ifndef HAVE_INODE_PERMISION_2ARGS
755 int ll_inode_permission(struct inode *inode, int mask, struct nameidata *nd);
757 int ll_inode_permission(struct inode *inode, int mask);
760 int ll_lov_setstripe_ea_info(struct inode *inode, struct file *file,
761 int flags, struct lov_user_md *lum,
763 int ll_lov_getstripe_ea_info(struct inode *inode, const char *filename,
764 struct lov_mds_md **lmm, int *lmm_size,
765 struct ptlrpc_request **request);
766 int ll_dir_setstripe(struct inode *inode, struct lov_user_md *lump,
768 int ll_dir_getstripe(struct inode *inode, struct lov_mds_md **lmm,
769 int *lmm_size, struct ptlrpc_request **request);
770 #ifdef HAVE_FILE_FSYNC_4ARGS
771 int ll_fsync(struct file *file, loff_t start, loff_t end, int data);
772 #elif defined(HAVE_FILE_FSYNC_2ARGS)
773 int ll_fsync(struct file *file, int data);
775 int ll_fsync(struct file *file, struct dentry *dentry, int data);
777 int ll_do_fiemap(struct inode *inode, struct ll_user_fiemap *fiemap,
779 int ll_merge_lvb(struct inode *inode);
780 int ll_get_grouplock(struct inode *inode, struct file *file, unsigned long arg);
781 int ll_put_grouplock(struct inode *inode, struct file *file, unsigned long arg);
782 int ll_fid2path(struct obd_export *exp, void *arg);
786 int ll_dops_init(struct dentry *de, int block, int init_sa);
787 extern struct dentry_operations ll_d_ops;
788 void ll_intent_drop_lock(struct lookup_intent *);
789 void ll_intent_release(struct lookup_intent *);
790 void ll_invalidate_aliases(struct inode *);
791 void ll_frob_intent(struct lookup_intent **itp, struct lookup_intent *deft);
792 void ll_lookup_finish_locks(struct lookup_intent *it, struct dentry *dentry);
793 #ifdef HAVE_D_COMPARE_7ARGS
794 int ll_dcompare(const struct dentry *parent, const struct inode *pinode,
795 const struct dentry *dentry, const struct inode *inode,
796 unsigned int len, const char *str, const struct qstr *d_name);
798 int ll_dcompare(struct dentry *parent, struct qstr *d_name, struct qstr *name);
800 int ll_revalidate_it_finish(struct ptlrpc_request *request,
801 struct lookup_intent *it, struct dentry *de);
803 /* llite/llite_lib.c */
804 extern struct super_operations lustre_super_operations;
806 char *ll_read_opt(const char *opt, char *data);
807 void ll_lli_init(struct ll_inode_info *lli);
808 int ll_fill_super(struct super_block *sb, struct vfsmount *mnt);
809 void ll_put_super(struct super_block *sb);
810 void ll_kill_super(struct super_block *sb);
811 struct inode *ll_inode_from_lock(struct ldlm_lock *lock);
812 void ll_clear_inode(struct inode *inode);
813 int ll_setattr_raw(struct dentry *dentry, struct iattr *attr);
814 int ll_setattr(struct dentry *de, struct iattr *attr);
815 int ll_statfs(struct dentry *de, struct kstatfs *sfs);
816 int ll_statfs_internal(struct super_block *sb, struct obd_statfs *osfs,
817 __u64 max_age, __u32 flags);
818 void ll_update_inode(struct inode *inode, struct lustre_md *md);
819 void ll_read_inode2(struct inode *inode, void *opaque);
820 void ll_delete_inode(struct inode *inode);
821 int ll_iocontrol(struct inode *inode, struct file *file,
822 unsigned int cmd, unsigned long arg);
823 int ll_flush_ctx(struct inode *inode);
824 #ifdef HAVE_UMOUNTBEGIN_VFSMOUNT
825 void ll_umount_begin(struct vfsmount *vfsmnt, int flags);
827 void ll_umount_begin(struct super_block *sb);
829 int ll_remount_fs(struct super_block *sb, int *flags, char *data);
830 int ll_show_options(struct seq_file *seq, struct vfsmount *vfs);
831 int ll_prep_inode(struct inode **inode, struct ptlrpc_request *req,
832 struct super_block *);
833 void lustre_dump_dentry(struct dentry *, int recur);
834 void lustre_dump_inode(struct inode *);
835 int ll_obd_statfs(struct inode *inode, void *arg);
836 int ll_get_max_mdsize(struct ll_sb_info *sbi, int *max_mdsize);
837 int ll_process_config(struct lustre_cfg *lcfg);
838 struct md_op_data *ll_prep_md_op_data(struct md_op_data *op_data,
839 struct inode *i1, struct inode *i2,
840 const char *name, int namelen,
841 int mode, __u32 opc, void *data);
842 void ll_finish_md_op_data(struct md_op_data *op_data);
843 int ll_get_obd_name(struct inode *inode, unsigned int cmd, unsigned long arg);
845 /* llite/llite_nfs.c */
846 extern struct export_operations lustre_export_operations;
847 __u32 get_uuid2int(const char *name, int len);
849 /* llite/special.c */
850 extern struct inode_operations ll_special_inode_operations;
851 extern struct file_operations ll_special_chr_inode_fops;
852 extern struct file_operations ll_special_chr_file_fops;
853 extern struct file_operations ll_special_blk_inode_fops;
854 extern struct file_operations ll_special_fifo_inode_fops;
855 extern struct file_operations ll_special_fifo_file_fops;
856 extern struct file_operations ll_special_sock_inode_fops;
858 /* llite/symlink.c */
859 extern struct inode_operations ll_fast_symlink_inode_operations;
861 /* llite/llite_close.c */
862 struct ll_close_queue {
863 cfs_spinlock_t lcq_lock;
865 cfs_waitq_t lcq_waitq;
866 cfs_completion_t lcq_comp;
867 cfs_atomic_t lcq_stop;
870 struct ccc_object *cl_inode2ccc(struct inode *inode);
873 void vvp_write_pending (struct ccc_object *club, struct ccc_page *page);
874 void vvp_write_complete(struct ccc_object *club, struct ccc_page *page);
876 /* specific achitecture can implement only part of this list */
877 enum vvp_io_subtype {
880 /** io called from .sendfile */
882 /** io started from splice_{read|write} */
889 enum vvp_io_subtype cui_io_subtype;
893 read_actor_t cui_actor;
897 struct pipe_inode_info *cui_pipe;
898 unsigned int cui_flags;
900 struct vvp_fault_io {
902 * Inode modification time that is checked across DLM
906 struct vm_area_struct *ft_vma;
908 * locked page returned from vvp_io
910 cfs_page_t *ft_vmpage;
911 #ifndef HAVE_VM_OP_FAULT
912 struct vm_nopage_api {
914 * Virtual address at which fault occurred.
916 unsigned long ft_address;
918 * Fault type, as to be supplied to
924 struct vm_fault_api {
928 struct vm_fault *ft_vmf;
930 * fault API used bitflags for return code.
932 unsigned int ft_flags;
938 * Read-ahead state used by read and page-fault IO contexts.
940 struct ll_ra_read cui_bead;
942 * Set when cui_bead has been initialized.
944 int cui_ra_window_set;
946 * Partially truncated page, that vvp_io_trunc_start() keeps locked
949 struct cl_page *cui_partpage;
953 * IO arguments for various VFS I/O interfaces.
956 /** normal/sendfile/splice */
957 enum vvp_io_subtype via_io_subtype;
961 #ifndef HAVE_FILE_WRITEV
962 struct kiocb *via_iocb;
964 struct iovec *via_iov;
965 unsigned long via_nrsegs;
968 read_actor_t via_actor;
972 struct pipe_inode_info *via_pipe;
973 unsigned int via_flags;
978 struct ll_cl_context {
980 struct cl_io *lcc_io;
981 struct cl_page *lcc_page;
982 struct lu_env *lcc_env;
987 struct vvp_thread_info {
988 struct ost_lvb vti_lvb;
989 struct cl_2queue vti_queue;
990 struct iovec vti_local_iov;
991 struct vvp_io_args vti_args;
992 struct ra_io_arg vti_ria;
993 struct kiocb vti_kiocb;
994 struct ll_cl_context vti_io_ctx;
997 static inline struct vvp_thread_info *vvp_env_info(const struct lu_env *env)
999 extern struct lu_context_key vvp_key;
1000 struct vvp_thread_info *info;
1002 info = lu_context_key_get(&env->le_ctx, &vvp_key);
1003 LASSERT(info != NULL);
1007 static inline struct vvp_io_args *vvp_env_args(const struct lu_env *env,
1008 enum vvp_io_subtype type)
1010 struct vvp_io_args *ret = &vvp_env_info(env)->vti_args;
1012 ret->via_io_subtype = type;
1017 struct vvp_session {
1018 struct vvp_io vs_ios;
1021 static inline struct vvp_session *vvp_env_session(const struct lu_env *env)
1023 extern struct lu_context_key vvp_session_key;
1024 struct vvp_session *ses;
1026 ses = lu_context_key_get(env->le_ses, &vvp_session_key);
1027 LASSERT(ses != NULL);
1031 static inline struct vvp_io *vvp_env_io(const struct lu_env *env)
1033 return &vvp_env_session(env)->vs_ios;
1036 void ll_queue_done_writing(struct inode *inode, unsigned long flags);
1037 void ll_close_thread_shutdown(struct ll_close_queue *lcq);
1038 int ll_close_thread_start(struct ll_close_queue **lcq_ret);
1040 /* llite/llite_mmap.c */
1041 typedef struct rb_root rb_root_t;
1042 typedef struct rb_node rb_node_t;
1044 struct ll_lock_tree_node;
1045 struct ll_lock_tree {
1047 cfs_list_t lt_locked_list;
1048 struct ll_file_data *lt_fd;
1051 int ll_teardown_mmaps(struct address_space *mapping, __u64 first, __u64 last);
1052 int ll_file_mmap(struct file * file, struct vm_area_struct * vma);
1053 struct ll_lock_tree_node * ll_node_from_inode(struct inode *inode, __u64 start,
1054 __u64 end, ldlm_mode_t mode);
1055 void policy_from_vma(ldlm_policy_data_t *policy,
1056 struct vm_area_struct *vma, unsigned long addr, size_t count);
1057 struct vm_area_struct *our_vma(struct mm_struct *mm, unsigned long addr,
1060 static inline void ll_invalidate_page(struct page *vmpage)
1062 struct address_space *mapping = vmpage->mapping;
1063 loff_t offset = vmpage->index << PAGE_CACHE_SHIFT;
1065 LASSERT(PageLocked(vmpage));
1066 if (mapping == NULL)
1069 ll_teardown_mmaps(mapping, offset, offset + CFS_PAGE_SIZE);
1070 truncate_complete_page(mapping, vmpage);
1073 #define ll_s2sbi(sb) (s2lsi(sb)->lsi_llsbi)
1075 /* don't need an addref as the sb_info should be holding one */
1076 static inline struct obd_export *ll_s2dtexp(struct super_block *sb)
1078 return ll_s2sbi(sb)->ll_dt_exp;
1081 /* don't need an addref as the sb_info should be holding one */
1082 static inline struct obd_export *ll_s2mdexp(struct super_block *sb)
1084 return ll_s2sbi(sb)->ll_md_exp;
1087 static inline struct client_obd *sbi2mdc(struct ll_sb_info *sbi)
1089 struct obd_device *obd = sbi->ll_md_exp->exp_obd;
1095 // FIXME: replace the name of this with LL_SB to conform to kernel stuff
1096 static inline struct ll_sb_info *ll_i2sbi(struct inode *inode)
1098 return ll_s2sbi(inode->i_sb);
1101 static inline struct obd_export *ll_i2dtexp(struct inode *inode)
1103 return ll_s2dtexp(inode->i_sb);
1106 static inline struct obd_export *ll_i2mdexp(struct inode *inode)
1108 return ll_s2mdexp(inode->i_sb);
1111 static inline struct lu_fid *ll_inode2fid(struct inode *inode)
1114 LASSERT(inode != NULL);
1115 fid = &ll_i2info(inode)->lli_fid;
1116 LASSERT(fid_is_igif(fid) || fid_ver(fid) == 0);
1120 static inline int ll_mds_max_easize(struct super_block *sb)
1122 return sbi2mdc(ll_s2sbi(sb))->cl_max_mds_easize;
1125 static inline __u64 ll_file_maxbytes(struct inode *inode)
1127 return ll_i2info(inode)->lli_maxbytes;
1131 int ll_setxattr(struct dentry *dentry, const char *name,
1132 const void *value, size_t size, int flags);
1133 ssize_t ll_getxattr(struct dentry *dentry, const char *name,
1134 void *buffer, size_t size);
1135 ssize_t ll_listxattr(struct dentry *dentry, char *buffer, size_t size);
1136 int ll_removexattr(struct dentry *dentry, const char *name);
1138 /* llite/remote_perm.c */
1139 extern cfs_mem_cache_t *ll_remote_perm_cachep;
1140 extern cfs_mem_cache_t *ll_rmtperm_hash_cachep;
1142 cfs_hlist_head_t *alloc_rmtperm_hash(void);
1143 void free_rmtperm_hash(cfs_hlist_head_t *hash);
1144 int ll_update_remote_perm(struct inode *inode, struct mdt_remote_perm *perm);
1145 int lustre_check_remote_perm(struct inode *inode, int mask);
1147 /* llite/llite_capa.c */
1148 extern cfs_timer_t ll_capa_timer;
1150 int ll_capa_thread_start(void);
1151 void ll_capa_thread_stop(void);
1152 void ll_capa_timer_callback(unsigned long unused);
1154 struct obd_capa *ll_add_capa(struct inode *inode, struct obd_capa *ocapa);
1155 int ll_update_capa(struct obd_capa *ocapa, struct lustre_capa *capa);
1157 void ll_capa_open(struct inode *inode);
1158 void ll_capa_close(struct inode *inode);
1160 struct obd_capa *ll_mdscapa_get(struct inode *inode);
1161 struct obd_capa *ll_osscapa_get(struct inode *inode, __u64 opc);
1163 void ll_truncate_free_capa(struct obd_capa *ocapa);
1164 void ll_clear_inode_capas(struct inode *inode);
1165 void ll_print_capa_stat(struct ll_sb_info *sbi);
1167 /* llite/llite_cl.c */
1168 extern struct lu_device_type vvp_device_type;
1171 * Common IO arguments for various VFS I/O interfaces.
1173 int cl_sb_init(struct super_block *sb);
1174 int cl_sb_fini(struct super_block *sb);
1175 enum cl_lock_mode vvp_mode_from_vma(struct vm_area_struct *vma);
1176 void ll_io_init(struct cl_io *io, const struct file *file, int write);
1178 void ras_update(struct ll_sb_info *sbi, struct inode *inode,
1179 struct ll_readahead_state *ras, unsigned long index,
1181 void ll_ra_count_put(struct ll_sb_info *sbi, unsigned long len);
1182 int ll_is_file_contended(struct file *file);
1183 void ll_ra_stats_inc(struct address_space *mapping, enum ra_stat which);
1185 /* llite/llite_rmtacl.c */
1186 #ifdef CONFIG_FS_POSIX_ACL
1187 obd_valid rce_ops2valid(int ops);
1188 struct rmtacl_ctl_entry *rct_search(struct rmtacl_ctl_table *rct, pid_t key);
1189 int rct_add(struct rmtacl_ctl_table *rct, pid_t key, int ops);
1190 int rct_del(struct rmtacl_ctl_table *rct, pid_t key);
1191 void rct_init(struct rmtacl_ctl_table *rct);
1192 void rct_fini(struct rmtacl_ctl_table *rct);
1194 void ee_free(struct eacl_entry *ee);
1195 int ee_add(struct eacl_table *et, pid_t key, struct lu_fid *fid, int type,
1196 ext_acl_xattr_header *header);
1197 struct eacl_entry *et_search_del(struct eacl_table *et, pid_t key,
1198 struct lu_fid *fid, int type);
1199 void et_search_free(struct eacl_table *et, pid_t key);
1200 void et_init(struct eacl_table *et);
1201 void et_fini(struct eacl_table *et);
1206 #define LL_SA_RPC_MIN 2
1207 #define LL_SA_RPC_DEF 32
1208 #define LL_SA_RPC_MAX 8192
1210 #define LL_SA_CACHE_BIT 5
1211 #define LL_SA_CACHE_SIZE (1 << LL_SA_CACHE_BIT)
1212 #define LL_SA_CACHE_MASK (LL_SA_CACHE_SIZE - 1)
1214 /* per inode struct, for dir only */
1215 struct ll_statahead_info {
1216 struct inode *sai_inode;
1217 cfs_atomic_t sai_refcount; /* when access this struct, hold
1219 unsigned int sai_generation; /* generation for statahead */
1220 unsigned int sai_max; /* max ahead of lookup */
1221 __u64 sai_sent; /* stat requests sent count */
1222 __u64 sai_replied; /* stat requests which received
1224 __u64 sai_index; /* index of statahead entry */
1225 __u64 sai_index_wait; /* index of entry which is the
1226 * caller is waiting for */
1227 __u64 sai_hit; /* hit count */
1228 __u64 sai_miss; /* miss count:
1229 * for "ls -al" case, it includes
1230 * hidden dentry miss;
1231 * for "ls -l" case, it does not
1232 * include hidden dentry miss.
1233 * "sai_miss_hidden" is used for
1236 unsigned int sai_consecutive_miss; /* consecutive miss */
1237 unsigned int sai_miss_hidden;/* "ls -al", but first dentry
1238 * is not a hidden one */
1239 unsigned int sai_skip_hidden;/* skipped hidden dentry count */
1240 unsigned int sai_ls_all:1, /* "ls -al", do stat-ahead for
1242 sai_in_readpage:1,/* statahead is in readdir()*/
1243 sai_agl_valid:1;/* AGL is valid for the dir */
1244 cfs_waitq_t sai_waitq; /* stat-ahead wait queue */
1245 struct ptlrpc_thread sai_thread; /* stat-ahead thread */
1246 struct ptlrpc_thread sai_agl_thread; /* AGL thread */
1247 cfs_list_t sai_entries_sent; /* entries sent out */
1248 cfs_list_t sai_entries_received; /* entries returned */
1249 cfs_list_t sai_entries_stated; /* entries stated */
1250 cfs_list_t sai_entries_agl; /* AGL entries to be sent */
1251 cfs_list_t sai_cache[LL_SA_CACHE_SIZE];
1252 cfs_spinlock_t sai_cache_lock[LL_SA_CACHE_SIZE];
1253 cfs_atomic_t sai_cache_count; /* entry count in cache */
1256 int do_statahead_enter(struct inode *dir, struct dentry **dentry,
1258 void ll_stop_statahead(struct inode *dir, void *key);
1260 static inline int ll_glimpse_size(struct inode *inode)
1262 struct ll_inode_info *lli = ll_i2info(inode);
1265 cfs_down_read(&lli->lli_glimpse_sem);
1266 rc = cl_glimpse_size(inode);
1267 lli->lli_glimpse_time = cfs_time_current();
1268 cfs_up_read(&lli->lli_glimpse_sem);
1273 ll_statahead_mark(struct inode *dir, struct dentry *dentry)
1275 struct ll_inode_info *lli = ll_i2info(dir);
1276 struct ll_statahead_info *sai = lli->lli_sai;
1277 struct ll_dentry_data *ldd = ll_d2d(dentry);
1279 /* not the same process, don't mark */
1280 if (lli->lli_opendir_pid != cfs_curproc_pid())
1283 if (sai != NULL && ldd != NULL)
1284 ldd->lld_sa_generation = sai->sai_generation;
1288 ll_statahead_enter(struct inode *dir, struct dentry **dentryp, int only_unplug)
1290 struct ll_inode_info *lli;
1291 struct ll_dentry_data *ldd;
1293 if (ll_i2sbi(dir)->ll_sa_max == 0)
1296 lli = ll_i2info(dir);
1297 /* not the same process, don't statahead */
1298 if (lli->lli_opendir_pid != cfs_curproc_pid())
1301 ldd = ll_d2d(*dentryp);
1303 * When stats a dentry, the system trigger more than once "revalidate"
1304 * or "lookup", for "getattr", for "getxattr", and maybe for others.
1305 * Under patchless client mode, the operation intent is not accurate,
1306 * which maybe misguide the statahead thread. For example:
1307 * The "revalidate" call for "getattr" and "getxattr" of a dentry maybe
1308 * have the same operation intent -- "IT_GETATTR".
1309 * In fact, one dentry should has only one chance to interact with the
1310 * statahead thread, otherwise the statahead windows will be confused.
1311 * The solution is as following:
1312 * Assign "lld_sa_generation" with "sai_generation" when a dentry
1313 * "IT_GETATTR" for the first time, and the subsequent "IT_GETATTR"
1314 * will bypass interacting with statahead thread for checking:
1315 * "lld_sa_generation == lli_sai->sai_generation"
1317 if (ldd && lli->lli_sai &&
1318 ldd->lld_sa_generation == lli->lli_sai->sai_generation)
1321 return do_statahead_enter(dir, dentryp, only_unplug);
1324 /* llite ioctl register support rountine */
1331 #define LLIOC_MAX_CMD 256
1334 * Rules to write a callback function:
1337 * @magic: Dynamic ioctl call routine will feed this vaule with the pointer
1338 * returned to ll_iocontrol_register. Callback functions should use this
1339 * data to check the potential collasion of ioctl cmd. If collasion is
1340 * found, callback function should return LLIOC_CONT.
1341 * @rcp: The result of ioctl command.
1344 * If @magic matches the pointer returned by ll_iocontrol_data, the
1345 * callback should return LLIOC_STOP; return LLIOC_STOP otherwise.
1347 typedef enum llioc_iter (*llioc_callback_t)(struct inode *inode,
1348 struct file *file, unsigned int cmd, unsigned long arg,
1349 void *magic, int *rcp);
1351 enum llioc_iter ll_iocontrol_call(struct inode *inode, struct file *file,
1352 unsigned int cmd, unsigned long arg, int *rcp);
1354 /* export functions */
1355 /* Register ioctl block dynamatically for a regular file.
1357 * @cmd: the array of ioctl command set
1358 * @count: number of commands in the @cmd
1359 * @cb: callback function, it will be called if an ioctl command is found to
1360 * belong to the command list @cmd.
1363 * A magic pointer will be returned if success;
1364 * otherwise, NULL will be returned.
1366 void *ll_iocontrol_register(llioc_callback_t cb, int count, unsigned int *cmd);
1367 void ll_iocontrol_unregister(void *magic);
1371 /* lclient compat stuff */
1372 #define cl_inode_info ll_inode_info
1373 #define cl_i2info(info) ll_i2info(info)
1374 #define cl_inode_mode(inode) ((inode)->i_mode)
1375 #define cl_i2sbi ll_i2sbi
1377 static inline struct ll_file_data *cl_iattr2fd(struct inode *inode,
1378 const struct iattr *attr)
1380 LASSERT(attr->ia_valid & ATTR_FILE);
1381 return LUSTRE_FPRIVATE(attr->ia_file);
1384 static inline void cl_isize_lock(struct inode *inode)
1386 ll_inode_size_lock(inode);
1389 static inline void cl_isize_unlock(struct inode *inode)
1391 ll_inode_size_unlock(inode);
1394 static inline void cl_isize_write_nolock(struct inode *inode, loff_t kms)
1396 LASSERT_SEM_LOCKED(&ll_i2info(inode)->lli_size_sem);
1397 i_size_write(inode, kms);
1400 static inline void cl_isize_write(struct inode *inode, loff_t kms)
1402 ll_inode_size_lock(inode);
1403 i_size_write(inode, kms);
1404 ll_inode_size_unlock(inode);
1407 #define cl_isize_read(inode) i_size_read(inode)
1409 static inline int cl_merge_lvb(struct inode *inode)
1411 return ll_merge_lvb(inode);
1414 #define cl_inode_atime(inode) LTIME_S((inode)->i_atime)
1415 #define cl_inode_ctime(inode) LTIME_S((inode)->i_ctime)
1416 #define cl_inode_mtime(inode) LTIME_S((inode)->i_mtime)
1418 struct obd_capa *cl_capa_lookup(struct inode *inode, enum cl_req_type crt);
1420 int cl_sync_file_range(struct inode *inode, loff_t start, loff_t end,
1421 enum cl_fsync_mode mode);
1423 /** direct write pages */
1424 struct ll_dio_pages {
1425 /** page array to be written. we don't support
1426 * partial pages except the last one. */
1427 struct page **ldp_pages;
1428 /* offset of each page */
1429 loff_t *ldp_offsets;
1430 /** if ldp_offsets is NULL, it means a sequential
1431 * pages to be written, then this is the file offset
1432 * of the * first page. */
1433 loff_t ldp_start_offset;
1434 /** how many bytes are to be written. */
1436 /** # of pages in the array. */
1440 static inline void cl_stats_tally(struct cl_device *dev, enum cl_req_type crt,
1443 int opc = (crt == CRT_READ) ? LPROC_LL_OSC_READ :
1446 ll_stats_ops_tally(ll_s2sbi(cl2ccc_dev(dev)->cdv_sb), opc, rc);
1449 extern ssize_t ll_direct_rw_pages(const struct lu_env *env, struct cl_io *io,
1450 int rw, struct inode *inode,
1451 struct ll_dio_pages *pv);
1453 static inline int ll_file_nolock(const struct file *file)
1455 struct ll_file_data *fd = LUSTRE_FPRIVATE(file);
1456 struct inode *inode = file->f_dentry->d_inode;
1458 LASSERT(fd != NULL);
1459 return ((fd->fd_flags & LL_FILE_IGNORE_LOCK) ||
1460 (ll_i2sbi(inode)->ll_flags & LL_SBI_NOLCK));
1463 static inline void ll_set_lock_data(struct obd_export *exp, struct inode *inode,
1464 struct lookup_intent *it, __u64 *bits)
1466 if (!it->d.lustre.it_lock_set) {
1467 CDEBUG(D_DLMTRACE, "setting l_data to inode %p (%lu/%u)\n",
1468 inode, inode->i_ino, inode->i_generation);
1469 md_set_lock_data(exp, &it->d.lustre.it_lock_handle,
1470 inode, &it->d.lustre.it_lock_bits);
1471 it->d.lustre.it_lock_set = 1;
1475 *bits = it->d.lustre.it_lock_bits;
1478 static inline void ll_lock_dcache(struct inode *inode)
1480 #ifdef HAVE_DCACHE_LOCK
1481 spin_lock(&dcache_lock);
1483 spin_lock(&inode->i_lock);
1487 static inline void ll_unlock_dcache(struct inode *inode)
1489 #ifdef HAVE_DCACHE_LOCK
1490 spin_unlock(&dcache_lock);
1492 spin_unlock(&inode->i_lock);
1496 static inline int d_lustre_invalid(const struct dentry *dentry)
1498 return dentry->d_flags & DCACHE_LUSTRE_INVALID;
1501 static inline void __d_lustre_invalidate(struct dentry *dentry)
1503 dentry->d_flags |= DCACHE_LUSTRE_INVALID;
1507 * Mark dentry INVALID, if dentry refcount is zero (this is normally case for
1508 * ll_md_blocking_ast), unhash this dentry, and let dcache to reclaim it later;
1509 * else dput() of the last refcount will unhash this dentry and kill it.
1511 static inline void d_lustre_invalidate(struct dentry *dentry)
1513 CDEBUG(D_DENTRY, "invalidate dentry %.*s (%p) parent %p inode %p "
1514 "refc %d\n", dentry->d_name.len, dentry->d_name.name, dentry,
1515 dentry->d_parent, dentry->d_inode, d_refcount(dentry));
1517 spin_lock(&dentry->d_lock);
1518 __d_lustre_invalidate(dentry);
1519 if (d_refcount(dentry) == 0)
1521 spin_unlock(&dentry->d_lock);
1524 static inline void d_lustre_revalidate(struct dentry *dentry)
1526 spin_lock(&dentry->d_lock);
1527 dentry->d_flags &= ~DCACHE_LUSTRE_INVALID;
1528 spin_unlock(&dentry->d_lock);
1531 #if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 7, 50, 0)
1532 /* Compatibility for old (1.8) compiled userspace quota code */
1533 struct if_quotactl_18 {
1538 struct obd_dqinfo qc_dqinfo;
1539 struct obd_dqblk qc_dqblk;
1541 struct obd_uuid obd_uuid;
1543 #define LL_IOC_QUOTACTL_18 _IOWR('f', 162, struct if_quotactl_18 *)
1544 /* End compatibility for old (1.8) compiled userspace quota code */
1546 #warning "remove old LL_IOC_QUOTACTL_18 compatibility code"
1547 #endif /* LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 7, 50, 0) */
1549 int ll_layout_conf(struct inode *inode, const struct cl_object_conf *conf);
1550 int ll_layout_refresh(struct inode *inode, __u32 *gen);
1552 #endif /* LLITE_INTERNAL_H */