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, Intel Corporation.
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
39 #include <lustre_debug.h>
40 #include <lustre_ver.h>
41 #include <lustre_disk.h> /* for s2sbi */
42 #include <lustre_eacl.h>
44 /* for struct cl_lock_descr and struct cl_io */
45 #include <cl_object.h>
47 #include <lustre_mdc.h>
48 #include <linux/lustre_intent.h>
49 #include <linux/compat.h>
50 #include <linux/posix_acl_xattr.h>
56 #ifndef VM_FAULT_RETRY
57 #define VM_FAULT_RETRY 0
60 /* Kernel 3.1 kills LOOKUP_CONTINUE, LOOKUP_PARENT is equivalent to it.
61 * seem kernel commit 49084c3bb2055c401f3493c13edae14d49128ca0 */
62 #ifndef LOOKUP_CONTINUE
63 #define LOOKUP_CONTINUE LOOKUP_PARENT
66 /** Only used on client-side for indicating the tail of dir hash/offset. */
67 #define LL_DIR_END_OFF 0x7fffffffffffffffULL
68 #define LL_DIR_END_OFF_32BIT 0x7fffffffUL
70 #define LL_IT2STR(it) ((it) ? ldlm_it2str((it)->it_op) : "0")
71 #define LUSTRE_FPRIVATE(file) ((file)->private_data)
73 struct ll_dentry_data
{
76 struct obd_client_handle lld_cwd_och
;
77 struct obd_client_handle lld_mnt_och
;
78 struct lookup_intent
*lld_it
;
79 unsigned int lld_sa_generation
;
80 unsigned int lld_invalid
:1;
81 struct rcu_head lld_rcu_head
;
84 #define ll_d2d(de) ((struct ll_dentry_data*)((de)->d_fsdata))
86 extern struct file_operations ll_pgcache_seq_fops
;
88 #define LLI_INODE_MAGIC 0x111d0de5
89 #define LLI_INODE_DEAD 0xdeadd00d
91 /* remote client permission cache */
92 #define REMOTE_PERM_HASHSIZE 16
94 struct ll_getname_data
{
95 struct dir_context ctx
;
96 char *lgd_name
; /* points to a buffer with NAME_MAX+1 size */
97 struct lu_fid lgd_fid
; /* target fid we are looking for */
98 int lgd_found
; /* inode matched? */
101 /* llite setxid/access permission for user on remote client */
102 struct ll_remote_perm
{
103 struct hlist_node lrp_list
;
108 int lrp_access_perm
; /* MAY_READ/WRITE/EXEC, this
109 is access permission with
110 lrp_fsuid/lrp_fsgid. */
114 /* MDS has an authority for the Size-on-MDS attributes. */
115 LLIF_MDS_SIZE_LOCK
= (1 << 0),
116 /* Epoch close is postponed. */
117 LLIF_EPOCH_PENDING
= (1 << 1),
118 /* DONE WRITING is allowed. */
119 LLIF_DONE_WRITING
= (1 << 2),
120 /* Sizeon-on-MDS attributes are changed. An attribute update needs to
122 LLIF_SOM_DIRTY
= (1 << 3),
123 /* File is contented */
124 LLIF_CONTENDED
= (1 << 4),
125 /* Truncate uses server lock for this file */
126 LLIF_SRVLOCK
= (1 << 5),
127 /* File data is modified. */
128 LLIF_DATA_MODIFIED
= (1 << 6),
129 /* File is being restored */
130 LLIF_FILE_RESTORING
= (1 << 7),
131 /* Xattr cache is attached to the file */
132 LLIF_XATTR_CACHE
= (1 << 8),
135 struct ll_inode_info
{
136 __u32 lli_inode_magic
;
141 struct posix_acl
*lli_posix_acl
;
143 struct hlist_head
*lli_remote_perms
;
144 struct mutex lli_rmtperm_mutex
;
146 /* identifying fields for both metadata and data stacks. */
147 struct lu_fid lli_fid
;
148 /* Parent fid for accessing default stripe data on parent directory
149 * for allocating OST objects after a mknod() and later open-by-FID. */
150 struct lu_fid lli_pfid
;
152 struct list_head lli_close_list
;
153 struct list_head lli_oss_capas
;
154 /* open count currently used by capability only, indicate whether
155 * capability needs renewal */
156 atomic_t lli_open_count
;
157 struct obd_capa
*lli_mds_capa
;
158 cfs_time_t lli_rmtperm_time
;
160 /* handle is to be sent to MDS later on done_writing and setattr.
161 * Open handle data are needed for the recovery to reconstruct
162 * the inode state on the MDS. XXX: recovery is not ready yet. */
163 struct obd_client_handle
*lli_pending_och
;
165 /* We need all three because every inode may be opened in different
167 struct obd_client_handle
*lli_mds_read_och
;
168 struct obd_client_handle
*lli_mds_write_och
;
169 struct obd_client_handle
*lli_mds_exec_och
;
170 __u64 lli_open_fd_read_count
;
171 __u64 lli_open_fd_write_count
;
172 __u64 lli_open_fd_exec_count
;
173 /* Protects access to och pointers and their usage counters */
174 struct mutex lli_och_mutex
;
176 struct inode lli_vfs_inode
;
178 /* the most recent timestamps obtained from mds */
179 struct ost_lvb lli_lvb
;
180 spinlock_t lli_agl_lock
;
182 /* Try to make the d::member and f::member are aligned. Before using
183 * these members, make clear whether it is directory or not. */
187 /* serialize normal readdir and statahead-readdir. */
188 struct mutex d_readdir_mutex
;
190 /* metadata statahead */
191 /* since parent-child threads can share the same @file
192 * struct, "opendir_key" is the token when dir close for
193 * case of parent exit before child -- it is me should
194 * cleanup the dir readahead. */
196 struct ll_statahead_info
*d_sai
;
197 struct posix_acl
*d_def_acl
;
198 /* protect statahead stuff. */
199 spinlock_t d_sa_lock
;
200 /* "opendir_pid" is the token when lookup/revalid
201 * -- I am the owner of dir statahead. */
205 #define lli_readdir_mutex u.d.d_readdir_mutex
206 #define lli_opendir_key u.d.d_opendir_key
207 #define lli_sai u.d.d_sai
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 struct semaphore f_size_sem
;
215 void *f_size_sem_owner
;
216 char *f_symlink_name
;
219 * struct rw_semaphore {
220 * signed long count; // align d.d_def_acl
221 * spinlock_t wait_lock; // align d.d_sa_lock
222 * struct list_head wait_list;
225 struct rw_semaphore f_trunc_sem
;
226 struct mutex f_write_mutex
;
228 struct rw_semaphore f_glimpse_sem
;
229 cfs_time_t f_glimpse_time
;
230 struct list_head f_agl_list
;
233 /* for writepage() only to communicate to fsync */
236 /* volatile file criteria is based on file name, this
237 * flag is used to keep the test result, so the strcmp
242 * whenever a process try to read/write the file, the
243 * jobid of the process will be saved here, and it'll
244 * be packed into the write PRC when flush later.
246 * so the read/write statistics for jobid will not be
247 * accurate if the file is shared by different jobs.
249 char f_jobid
[JOBSTATS_JOBID_SIZE
];
252 #define lli_size_sem u.f.f_size_sem
253 #define lli_size_sem_owner u.f.f_size_sem_owner
254 #define lli_symlink_name u.f.f_symlink_name
255 #define lli_maxbytes u.f.f_maxbytes
256 #define lli_trunc_sem u.f.f_trunc_sem
257 #define lli_write_mutex u.f.f_write_mutex
258 #define lli_glimpse_sem u.f.f_glimpse_sem
259 #define lli_glimpse_time u.f.f_glimpse_time
260 #define lli_agl_list u.f.f_agl_list
261 #define lli_agl_index u.f.f_agl_index
262 #define lli_async_rc u.f.f_async_rc
263 #define lli_jobid u.f.f_jobid
264 #define lli_volatile u.f.f_volatile
268 /* XXX: For following frequent used members, although they maybe special
269 * used for non-directory object, it is some time-wasting to check
270 * whether the object is directory or not before using them. On the
271 * other hand, currently, sizeof(f) > sizeof(d), it cannot reduce
272 * the "ll_inode_info" size even if moving those members into u.f.
273 * So keep them out side.
275 * In the future, if more members are added only for directory,
276 * some of the following members can be moved into u.f.
279 struct cl_object
*lli_clob
;
281 /* mutex to request for layout lock exclusively. */
282 struct mutex lli_layout_mutex
;
283 /* valid only inside LAYOUT ibits lock, protected by lli_layout_mutex */
284 __u32 lli_layout_gen
;
286 struct rw_semaphore lli_xattrs_list_rwsem
;
287 struct mutex lli_xattrs_enq_lock
;
288 struct list_head lli_xattrs
;/* ll_xattr_entry->xe_list */
291 int ll_xattr_cache_destroy(struct inode
*inode
);
293 int ll_xattr_cache_get(struct inode
*inode
,
299 int ll_xattr_cache_update(struct inode
*inode
,
307 * Locking to guarantee consistency of non-atomic updates to long long i_size,
308 * consistency between file size and KMS.
310 * Implemented by ->lli_size_sem and ->lsm_lock, nested in that order.
313 void ll_inode_size_lock(struct inode
*inode
);
314 void ll_inode_size_unlock(struct inode
*inode
);
316 // FIXME: replace the name of this with LL_I to conform to kernel stuff
317 // static inline struct ll_inode_info *LL_I(struct inode *inode)
318 static inline struct ll_inode_info
*ll_i2info(struct inode
*inode
)
320 return container_of(inode
, struct ll_inode_info
, lli_vfs_inode
);
323 /* default to about 40meg of readahead on a given system. That much tied
324 * up in 512k readahead requests serviced at 40ms each is about 1GB/s. */
325 #define SBI_DEFAULT_READAHEAD_MAX (40UL << (20 - PAGE_CACHE_SHIFT))
327 /* default to read-ahead full files smaller than 2MB on the second read */
328 #define SBI_DEFAULT_READAHEAD_WHOLE_MAX (2UL << (20 - PAGE_CACHE_SHIFT))
333 RA_STAT_DISTANT_READPAGE
,
334 RA_STAT_MISS_IN_WINDOW
,
335 RA_STAT_FAILED_GRAB_PAGE
,
336 RA_STAT_FAILED_MATCH
,
341 RA_STAT_MAX_IN_FLIGHT
,
342 RA_STAT_WRONG_GRAB_PAGE
,
347 atomic_t ra_cur_pages
;
348 unsigned long ra_max_pages
;
349 unsigned long ra_max_pages_per_file
;
350 unsigned long ra_max_read_ahead_whole_pages
;
353 /* ra_io_arg will be filled in the beginning of ll_readahead with
354 * ras_lock, then the following ll_read_ahead_pages will read RA
355 * pages according to this arg, all the items in this structure are
356 * counted by page index.
359 unsigned long ria_start
; /* start offset of read-ahead*/
360 unsigned long ria_end
; /* end offset of read-ahead*/
361 /* If stride read pattern is detected, ria_stoff means where
362 * stride read is started. Note: for normal read-ahead, the
363 * value here is meaningless, and also it will not be accessed*/
365 /* ria_length and ria_pages are the length and pages length in the
366 * stride I/O mode. And they will also be used to check whether
367 * it is stride I/O read-ahead in the read-ahead pages*/
368 unsigned long ria_length
;
369 unsigned long ria_pages
;
372 /* LL_HIST_MAX=32 causes an overflow */
373 #define LL_HIST_MAX 28
374 #define LL_HIST_START 12 /* buckets start at 2^12 = 4k */
375 #define LL_PROCESS_HIST_MAX 10
376 struct per_process_info
{
378 struct obd_histogram pp_r_hist
;
379 struct obd_histogram pp_w_hist
;
382 /* pp_extents[LL_PROCESS_HIST_MAX] will hold the combined process info */
383 struct ll_rw_extents_info
{
384 struct per_process_info pp_extents
[LL_PROCESS_HIST_MAX
+ 1];
387 #define LL_OFFSET_HIST_MAX 100
388 struct ll_rw_process_info
{
391 loff_t rw_range_start
;
393 loff_t rw_last_file_pos
;
395 size_t rw_smallest_extent
;
396 size_t rw_largest_extent
;
397 struct ll_file_data
*rw_last_file
;
400 enum stats_track_type
{
401 STATS_TRACK_ALL
= 0, /* track all processes */
402 STATS_TRACK_PID
, /* track process with this pid */
403 STATS_TRACK_PPID
, /* track processes with this ppid */
404 STATS_TRACK_GID
, /* track processes with this gid */
408 /* flags for sbi->ll_flags */
409 #define LL_SBI_NOLCK 0x01 /* DLM locking disabled (directio-only) */
410 #define LL_SBI_CHECKSUM 0x02 /* checksum each page as it's written */
411 #define LL_SBI_FLOCK 0x04
412 #define LL_SBI_USER_XATTR 0x08 /* support user xattr */
413 #define LL_SBI_ACL 0x10 /* support ACL */
414 #define LL_SBI_RMT_CLIENT 0x40 /* remote client */
415 #define LL_SBI_MDS_CAPA 0x80 /* support mds capa */
416 #define LL_SBI_OSS_CAPA 0x100 /* support oss capa */
417 #define LL_SBI_LOCALFLOCK 0x200 /* Local flocks support by kernel */
418 #define LL_SBI_LRU_RESIZE 0x400 /* lru resize support */
419 #define LL_SBI_LAZYSTATFS 0x800 /* lazystatfs mount option */
420 #define LL_SBI_SOM_PREVIEW 0x1000 /* SOM preview mount option */
421 #define LL_SBI_32BIT_API 0x2000 /* generate 32 bit inodes. */
422 #define LL_SBI_64BIT_HASH 0x4000 /* support 64-bits dir hash/offset */
423 #define LL_SBI_AGL_ENABLED 0x8000 /* enable agl */
424 #define LL_SBI_VERBOSE 0x10000 /* verbose mount/umount */
425 #define LL_SBI_LAYOUT_LOCK 0x20000 /* layout lock support */
426 #define LL_SBI_USER_FID2PATH 0x40000 /* allow fid2path by unprivileged users */
427 #define LL_SBI_XATTR_CACHE 0x80000 /* support for xattr cache */
429 #define LL_SBI_FLAGS { \
452 /* default value for ll_sb_info->contention_time */
453 #define SBI_DEFAULT_CONTENTION_SECONDS 60
454 /* default value for lockless_truncate_enable */
455 #define SBI_DEFAULT_LOCKLESS_TRUNCATE_ENABLE 1
456 #define RCE_HASHES 32
458 struct rmtacl_ctl_entry
{
459 struct list_head rce_list
;
460 pid_t rce_key
; /* hash key */
461 int rce_ops
; /* acl operation type */
464 struct rmtacl_ctl_table
{
466 struct list_head rct_entries
[RCE_HASHES
];
473 struct list_head et_entries
[EE_HASHES
];
477 struct list_head ll_list
;
478 /* this protects pglist and ra_info. It isn't safe to
479 * grab from interrupt contexts */
481 spinlock_t ll_pp_extent_lock
; /* pp_extent entry*/
482 spinlock_t ll_process_lock
; /* ll_rw_process_info */
483 struct obd_uuid ll_sb_uuid
;
484 struct obd_export
*ll_md_exp
;
485 struct obd_export
*ll_dt_exp
;
486 struct proc_dir_entry
* ll_proc_root
;
487 struct lu_fid ll_root_fid
; /* root object fid */
490 unsigned int ll_umounting
:1,
491 ll_xattr_cache_enabled
:1;
492 struct list_head ll_conn_chain
; /* per-conn chain of SBs */
493 struct lustre_client_ocd ll_lco
;
495 struct list_head ll_orphan_dentry_list
; /*please don't ask -p*/
496 struct ll_close_queue
*ll_lcq
;
498 struct lprocfs_stats
*ll_stats
; /* lprocfs stats counter */
500 struct cl_client_cache ll_cache
;
502 struct lprocfs_stats
*ll_ra_stats
;
504 struct ll_ra_info ll_ra_info
;
505 unsigned int ll_namelen
;
506 struct file_operations
*ll_fop
;
508 /* =0 - hold lock over whole read/write
509 * >0 - max. chunk to be read/written w/o lock re-acquiring */
510 unsigned long ll_max_rw_chunk
;
511 unsigned int ll_md_brw_size
; /* used by readdir */
513 struct lu_site
*ll_site
;
514 struct cl_device
*ll_cl
;
516 struct ll_rw_extents_info ll_rw_extents_info
;
517 int ll_extent_process_count
;
518 struct ll_rw_process_info ll_rw_process_info
[LL_PROCESS_HIST_MAX
];
519 unsigned int ll_offset_process_count
;
520 struct ll_rw_process_info ll_rw_offset_info
[LL_OFFSET_HIST_MAX
];
521 unsigned int ll_rw_offset_entry_count
;
522 int ll_stats_track_id
;
523 enum stats_track_type ll_stats_track_type
;
526 /* metadata stat-ahead */
527 unsigned int ll_sa_max
; /* max statahead RPCs */
528 atomic_t ll_sa_total
; /* statahead thread started
530 atomic_t ll_sa_wrong
; /* statahead thread stopped for
532 atomic_t ll_agl_total
; /* AGL thread started count */
534 dev_t ll_sdev_orig
; /* save s_dev before assign for
536 struct rmtacl_ctl_table ll_rct
;
537 struct eacl_table ll_et
;
538 __kernel_fsid_t ll_fsid
;
541 #define LL_DEFAULT_MAX_RW_CHUNK (32 * 1024 * 1024)
546 struct task_struct
*lrr_reader
;
547 struct list_head lrr_linkage
;
551 * per file-descriptor read-ahead data.
553 struct ll_readahead_state
{
556 * index of the last page that read(2) needed and that wasn't in the
557 * cache. Used by ras_update() to detect seeks.
559 * XXX nikita: if access seeks into cached region, Lustre doesn't see
562 unsigned long ras_last_readpage
;
564 * number of pages read after last read-ahead window reset. As window
565 * is reset on each seek, this is effectively a number of consecutive
566 * accesses. Maybe ->ras_accessed_in_window is better name.
568 * XXX nikita: window is also reset (by ras_update()) when Lustre
569 * believes that memory pressure evicts read-ahead pages. In that
570 * case, it probably doesn't make sense to expand window to
571 * PTLRPC_MAX_BRW_PAGES on the third access.
573 unsigned long ras_consecutive_pages
;
575 * number of read requests after the last read-ahead window reset
576 * As window is reset on each seek, this is effectively the number
577 * on consecutive read request and is used to trigger read-ahead.
579 unsigned long ras_consecutive_requests
;
581 * Parameters of current read-ahead window. Handled by
582 * ras_update(). On the initial access to the file or after a seek,
583 * window is reset to 0. After 3 consecutive accesses, window is
584 * expanded to PTLRPC_MAX_BRW_PAGES. Afterwards, window is enlarged by
585 * PTLRPC_MAX_BRW_PAGES chunks up to ->ra_max_pages.
587 unsigned long ras_window_start
, ras_window_len
;
589 * Where next read-ahead should start at. This lies within read-ahead
590 * window. Read-ahead window is read in pieces rather than at once
591 * because: 1. lustre limits total number of pages under read-ahead by
592 * ->ra_max_pages (see ll_ra_count_get()), 2. client cannot read pages
593 * not covered by DLM lock.
595 unsigned long ras_next_readahead
;
597 * Total number of ll_file_read requests issued, reads originating
598 * due to mmap are not counted in this total. This value is used to
599 * trigger full file read-ahead after multiple reads to a small file.
601 unsigned long ras_requests
;
603 * Page index with respect to the current request, these value
604 * will not be accurate when dealing with reads issued via mmap.
606 unsigned long ras_request_index
;
608 * list of struct ll_ra_read's one per read(2) call current in
609 * progress against this file descriptor. Used by read-ahead code,
610 * protected by ->ras_lock.
612 struct list_head ras_read_beads
;
614 * The following 3 items are used for detecting the stride I/O
616 * In stride I/O mode,
617 * ...............|-----data-----|****gap*****|--------|******|....
618 * offset |-stride_pages-|-stride_gap-|
619 * ras_stride_offset = offset;
620 * ras_stride_length = stride_pages + stride_gap;
621 * ras_stride_pages = stride_pages;
622 * Note: all these three items are counted by pages.
624 unsigned long ras_stride_length
;
625 unsigned long ras_stride_pages
;
626 pgoff_t ras_stride_offset
;
628 * number of consecutive stride request count, and it is similar as
629 * ras_consecutive_requests, but used for stride I/O mode.
630 * Note: only more than 2 consecutive stride request are detected,
631 * stride read-ahead will be enable
633 unsigned long ras_consecutive_stride_requests
;
636 extern struct kmem_cache
*ll_file_data_slab
;
637 struct lustre_handle
;
638 struct ll_file_data
{
639 struct ll_readahead_state fd_ras
;
640 struct ccc_grouplock fd_grouplock
;
644 /* openhandle if lease exists for this file.
645 * Borrow lli->lli_och_mutex to protect assignment */
646 struct obd_client_handle
*fd_lease_och
;
647 struct obd_client_handle
*fd_och
;
648 struct file
*fd_file
;
649 /* Indicate whether need to report failure when close.
650 * true: failure is known, not report again.
651 * false: unknown failure, should report. */
652 bool fd_write_failed
;
655 struct lov_stripe_md
;
657 extern spinlock_t inode_lock
;
659 extern struct proc_dir_entry
*proc_lustre_fs_root
;
661 static inline struct inode
*ll_info2i(struct ll_inode_info
*lli
)
663 return &lli
->lli_vfs_inode
;
667 struct inode
*icbd_parent
;
668 struct dentry
**icbd_childp
;
672 __u32
ll_i2suppgid(struct inode
*i
);
673 void ll_i2gids(__u32
*suppgids
, struct inode
*i1
,struct inode
*i2
);
675 static inline int ll_need_32bit_api(struct ll_sb_info
*sbi
)
677 #if BITS_PER_LONG == 32
684 (sbi
->ll_flags
& LL_SBI_32BIT_API
)
689 #define LLAP_MAGIC 98764321
691 extern struct kmem_cache
*ll_async_page_slab
;
692 extern size_t ll_async_page_slab_size
;
694 void ll_ra_read_in(struct file
*f
, struct ll_ra_read
*rar
);
695 void ll_ra_read_ex(struct file
*f
, struct ll_ra_read
*rar
);
696 struct ll_ra_read
*ll_ra_read_get(struct file
*f
);
698 /* llite/lproc_llite.c */
700 int lprocfs_register_mountpoint(struct proc_dir_entry
*parent
,
701 struct super_block
*sb
, char *osc
, char *mdc
);
702 void lprocfs_unregister_mountpoint(struct ll_sb_info
*sbi
);
703 void ll_stats_ops_tally(struct ll_sb_info
*sbi
, int op
, int count
);
704 void lprocfs_llite_init_vars(struct lprocfs_static_vars
*lvars
);
705 void ll_rw_stats_tally(struct ll_sb_info
*sbi
, pid_t pid
,
706 struct ll_file_data
*file
, loff_t pos
,
707 size_t count
, int rw
);
709 static inline int lprocfs_register_mountpoint(struct proc_dir_entry
*parent
,
710 struct super_block
*sb
, char *osc
, char *mdc
){return 0;}
711 static inline void lprocfs_unregister_mountpoint(struct ll_sb_info
*sbi
) {}
713 void ll_stats_ops_tally(struct ll_sb_info
*sbi
, int op
, int count
) {}
714 static inline void lprocfs_llite_init_vars(struct lprocfs_static_vars
*lvars
)
716 memset(lvars
, 0, sizeof(*lvars
));
718 static inline void ll_rw_stats_tally(struct ll_sb_info
*sbi
, pid_t pid
,
719 struct ll_file_data
*file
, loff_t pos
,
720 size_t count
, int rw
) {}
725 void ll_release_page(struct page
*page
, int remove
);
726 extern struct file_operations ll_dir_operations
;
727 extern struct inode_operations ll_dir_inode_operations
;
728 struct page
*ll_get_dir_page(struct inode
*dir
, __u64 hash
,
729 struct ll_dir_chain
*chain
);
730 int ll_dir_read(struct inode
*inode
, struct dir_context
*ctx
);
732 int ll_get_mdt_idx(struct inode
*inode
);
734 int ll_objects_destroy(struct ptlrpc_request
*request
,
736 struct inode
*ll_iget(struct super_block
*sb
, ino_t hash
,
737 struct lustre_md
*lic
);
738 int ll_md_blocking_ast(struct ldlm_lock
*, struct ldlm_lock_desc
*,
739 void *data
, int flag
);
740 struct dentry
*ll_splice_alias(struct inode
*inode
, struct dentry
*de
);
741 int ll_rmdir_entry(struct inode
*dir
, char *name
, int namelen
);
744 int ll_prepare_write(struct file
*, struct page
*, unsigned from
, unsigned to
);
745 int ll_commit_write(struct file
*, struct page
*, unsigned from
, unsigned to
);
746 int ll_writepage(struct page
*page
, struct writeback_control
*wbc
);
747 int ll_writepages(struct address_space
*, struct writeback_control
*wbc
);
748 void ll_removepage(struct page
*page
);
749 int ll_readpage(struct file
*file
, struct page
*page
);
750 void ll_readahead_init(struct inode
*inode
, struct ll_readahead_state
*ras
);
751 int ll_file_punch(struct inode
*, loff_t
, int);
752 ssize_t
ll_file_lockless_io(struct file
*, char *, size_t, loff_t
*, int);
753 void ll_clear_file_contended(struct inode
*);
754 int ll_sync_page_range(struct inode
*, struct address_space
*, loff_t
, size_t);
755 int ll_readahead(const struct lu_env
*env
, struct cl_io
*io
,
756 struct ll_readahead_state
*ras
, struct address_space
*mapping
,
757 struct cl_page_list
*queue
, int flags
);
760 extern struct file_operations ll_file_operations
;
761 extern struct file_operations ll_file_operations_flock
;
762 extern struct file_operations ll_file_operations_noflock
;
763 extern struct inode_operations ll_file_inode_operations
;
764 extern int ll_inode_revalidate_it(struct dentry
*, struct lookup_intent
*,
766 extern int ll_have_md_lock(struct inode
*inode
, __u64
*bits
,
767 ldlm_mode_t l_req_mode
);
768 extern ldlm_mode_t
ll_take_md_lock(struct inode
*inode
, __u64 bits
,
769 struct lustre_handle
*lockh
, __u64 flags
,
771 int __ll_inode_revalidate_it(struct dentry
*, struct lookup_intent
*,
773 int ll_revalidate_nd(struct dentry
*dentry
, unsigned int flags
);
774 int ll_file_open(struct inode
*inode
, struct file
*file
);
775 int ll_file_release(struct inode
*inode
, struct file
*file
);
776 int ll_glimpse_ioctl(struct ll_sb_info
*sbi
,
777 struct lov_stripe_md
*lsm
, lstat_t
*st
);
778 void ll_ioepoch_open(struct ll_inode_info
*lli
, __u64 ioepoch
);
779 int ll_local_open(struct file
*file
,
780 struct lookup_intent
*it
, struct ll_file_data
*fd
,
781 struct obd_client_handle
*och
);
782 int ll_release_openhandle(struct dentry
*, struct lookup_intent
*);
783 int ll_md_close(struct obd_export
*md_exp
, struct inode
*inode
,
785 int ll_md_real_close(struct inode
*inode
, int flags
);
786 void ll_ioepoch_close(struct inode
*inode
, struct md_op_data
*op_data
,
787 struct obd_client_handle
**och
, unsigned long flags
);
788 void ll_done_writing_attr(struct inode
*inode
, struct md_op_data
*op_data
);
789 int ll_som_update(struct inode
*inode
, struct md_op_data
*op_data
);
790 int ll_inode_getattr(struct inode
*inode
, struct obdo
*obdo
,
791 __u64 ioepoch
, int sync
);
792 int ll_md_setattr(struct dentry
*dentry
, struct md_op_data
*op_data
,
793 struct md_open_data
**mod
);
794 void ll_pack_inode2opdata(struct inode
*inode
, struct md_op_data
*op_data
,
795 struct lustre_handle
*fh
);
796 int ll_getattr_it(struct vfsmount
*mnt
, struct dentry
*de
,
797 struct lookup_intent
*it
, struct kstat
*stat
);
798 int ll_getattr(struct vfsmount
*mnt
, struct dentry
*de
, struct kstat
*stat
);
799 struct ll_file_data
*ll_file_data_get(void);
800 struct posix_acl
* ll_get_acl(struct inode
*inode
, int type
);
802 int ll_inode_permission(struct inode
*inode
, int mask
);
804 int ll_lov_setstripe_ea_info(struct inode
*inode
, struct file
*file
,
805 int flags
, struct lov_user_md
*lum
,
807 int ll_lov_getstripe_ea_info(struct inode
*inode
, const char *filename
,
808 struct lov_mds_md
**lmm
, int *lmm_size
,
809 struct ptlrpc_request
**request
);
810 int ll_dir_setstripe(struct inode
*inode
, struct lov_user_md
*lump
,
812 int ll_dir_getstripe(struct inode
*inode
, struct lov_mds_md
**lmmp
,
813 int *lmm_size
, struct ptlrpc_request
**request
);
814 int ll_fsync(struct file
*file
, loff_t start
, loff_t end
, int data
);
815 int ll_do_fiemap(struct inode
*inode
, struct ll_user_fiemap
*fiemap
,
817 int ll_merge_lvb(const struct lu_env
*env
, struct inode
*inode
);
818 int ll_get_grouplock(struct inode
*inode
, struct file
*file
, unsigned long arg
);
819 int ll_put_grouplock(struct inode
*inode
, struct file
*file
, unsigned long arg
);
820 int ll_fid2path(struct inode
*inode
, void *arg
);
821 int ll_data_version(struct inode
*inode
, __u64
*data_version
, int extent_lock
);
822 int ll_hsm_release(struct inode
*inode
);
824 struct obd_client_handle
*ll_lease_open(struct inode
*inode
, struct file
*file
,
825 fmode_t mode
, __u64 flags
);
826 int ll_lease_close(struct obd_client_handle
*och
, struct inode
*inode
,
831 int ll_d_init(struct dentry
*de
);
832 extern struct dentry_operations ll_d_ops
;
833 void ll_intent_drop_lock(struct lookup_intent
*);
834 void ll_intent_release(struct lookup_intent
*);
835 void ll_invalidate_aliases(struct inode
*);
836 void ll_frob_intent(struct lookup_intent
**itp
, struct lookup_intent
*deft
);
837 void ll_lookup_finish_locks(struct lookup_intent
*it
, struct dentry
*dentry
);
838 int ll_dcompare(const struct dentry
*parent
, const struct dentry
*dentry
,
839 unsigned int len
, const char *str
, const struct qstr
*d_name
);
840 int ll_revalidate_it_finish(struct ptlrpc_request
*request
,
841 struct lookup_intent
*it
, struct dentry
*de
);
843 /* llite/llite_lib.c */
844 extern struct super_operations lustre_super_operations
;
846 char *ll_read_opt(const char *opt
, char *data
);
847 void ll_lli_init(struct ll_inode_info
*lli
);
848 int ll_fill_super(struct super_block
*sb
, struct vfsmount
*mnt
);
849 void ll_put_super(struct super_block
*sb
);
850 void ll_kill_super(struct super_block
*sb
);
851 struct inode
*ll_inode_from_resource_lock(struct ldlm_lock
*lock
);
852 struct inode
*ll_inode_from_lock(struct ldlm_lock
*lock
);
853 void ll_clear_inode(struct inode
*inode
);
854 int ll_setattr_raw(struct dentry
*dentry
, struct iattr
*attr
, bool hsm_import
);
855 int ll_setattr(struct dentry
*de
, struct iattr
*attr
);
856 int ll_statfs(struct dentry
*de
, struct kstatfs
*sfs
);
857 int ll_statfs_internal(struct super_block
*sb
, struct obd_statfs
*osfs
,
858 __u64 max_age
, __u32 flags
);
859 void ll_update_inode(struct inode
*inode
, struct lustre_md
*md
);
860 void ll_read_inode2(struct inode
*inode
, void *opaque
);
861 void ll_delete_inode(struct inode
*inode
);
862 int ll_iocontrol(struct inode
*inode
, struct file
*file
,
863 unsigned int cmd
, unsigned long arg
);
864 int ll_flush_ctx(struct inode
*inode
);
865 void ll_umount_begin(struct super_block
*sb
);
866 int ll_remount_fs(struct super_block
*sb
, int *flags
, char *data
);
867 int ll_show_options(struct seq_file
*seq
, struct dentry
*dentry
);
868 void ll_dirty_page_discard_warn(struct page
*page
, int ioret
);
869 int ll_prep_inode(struct inode
**inode
, struct ptlrpc_request
*req
,
870 struct super_block
*, struct lookup_intent
*);
871 void lustre_dump_dentry(struct dentry
*, int recur
);
872 void lustre_dump_inode(struct inode
*);
873 int ll_obd_statfs(struct inode
*inode
, void *arg
);
874 int ll_get_max_mdsize(struct ll_sb_info
*sbi
, int *max_mdsize
);
875 int ll_process_config(struct lustre_cfg
*lcfg
);
876 struct md_op_data
*ll_prep_md_op_data(struct md_op_data
*op_data
,
877 struct inode
*i1
, struct inode
*i2
,
878 const char *name
, int namelen
,
879 int mode
, __u32 opc
, void *data
);
880 void ll_finish_md_op_data(struct md_op_data
*op_data
);
881 int ll_get_obd_name(struct inode
*inode
, unsigned int cmd
, unsigned long arg
);
882 char *ll_get_fsname(struct super_block
*sb
, char *buf
, int buflen
);
884 /* llite/llite_nfs.c */
885 extern struct export_operations lustre_export_operations
;
886 __u32
get_uuid2int(const char *name
, int len
);
887 void get_uuid2fsid(const char *name
, int len
, __kernel_fsid_t
*fsid
);
888 struct inode
*search_inode_for_lustre(struct super_block
*sb
,
889 const struct lu_fid
*fid
);
891 /* llite/special.c */
892 extern struct inode_operations ll_special_inode_operations
;
893 extern struct file_operations ll_special_chr_inode_fops
;
894 extern struct file_operations ll_special_chr_file_fops
;
895 extern struct file_operations ll_special_blk_inode_fops
;
896 extern struct file_operations ll_special_fifo_inode_fops
;
897 extern struct file_operations ll_special_fifo_file_fops
;
898 extern struct file_operations ll_special_sock_inode_fops
;
900 /* llite/symlink.c */
901 extern struct inode_operations ll_fast_symlink_inode_operations
;
903 /* llite/llite_close.c */
904 struct ll_close_queue
{
906 struct list_head lcq_head
;
907 wait_queue_head_t lcq_waitq
;
908 struct completion lcq_comp
;
912 struct ccc_object
*cl_inode2ccc(struct inode
*inode
);
915 void vvp_write_pending (struct ccc_object
*club
, struct ccc_page
*page
);
916 void vvp_write_complete(struct ccc_object
*club
, struct ccc_page
*page
);
918 /* specific achitecture can implement only part of this list */
919 enum vvp_io_subtype
{
922 /** io called from .sendfile */
924 /** io started from splice_{read|write} */
931 enum vvp_io_subtype cui_io_subtype
;
935 read_actor_t cui_actor
;
939 struct pipe_inode_info
*cui_pipe
;
940 unsigned int cui_flags
;
942 struct vvp_fault_io
{
944 * Inode modification time that is checked across DLM
948 struct vm_area_struct
*ft_vma
;
950 * locked page returned from vvp_io
952 struct page
*ft_vmpage
;
953 struct vm_fault_api
{
957 struct vm_fault
*ft_vmf
;
959 * fault API used bitflags for return code.
961 unsigned int ft_flags
;
966 * Read-ahead state used by read and page-fault IO contexts.
968 struct ll_ra_read cui_bead
;
970 * Set when cui_bead has been initialized.
972 int cui_ra_window_set
;
974 * Partially truncated page, that vvp_io_trunc_start() keeps locked
977 struct cl_page
*cui_partpage
;
981 * IO arguments for various VFS I/O interfaces.
984 /** normal/sendfile/splice */
985 enum vvp_io_subtype via_io_subtype
;
989 struct kiocb
*via_iocb
;
990 struct iovec
*via_iov
;
991 unsigned long via_nrsegs
;
994 read_actor_t via_actor
;
998 struct pipe_inode_info
*via_pipe
;
999 unsigned int via_flags
;
1004 struct ll_cl_context
{
1006 struct cl_io
*lcc_io
;
1007 struct cl_page
*lcc_page
;
1008 struct lu_env
*lcc_env
;
1013 struct vvp_thread_info
{
1014 struct ost_lvb vti_lvb
;
1015 struct cl_2queue vti_queue
;
1016 struct iovec vti_local_iov
;
1017 struct vvp_io_args vti_args
;
1018 struct ra_io_arg vti_ria
;
1019 struct kiocb vti_kiocb
;
1020 struct ll_cl_context vti_io_ctx
;
1023 static inline struct vvp_thread_info
*vvp_env_info(const struct lu_env
*env
)
1025 extern struct lu_context_key vvp_key
;
1026 struct vvp_thread_info
*info
;
1028 info
= lu_context_key_get(&env
->le_ctx
, &vvp_key
);
1029 LASSERT(info
!= NULL
);
1033 static inline struct vvp_io_args
*vvp_env_args(const struct lu_env
*env
,
1034 enum vvp_io_subtype type
)
1036 struct vvp_io_args
*ret
= &vvp_env_info(env
)->vti_args
;
1038 ret
->via_io_subtype
= type
;
1043 struct vvp_session
{
1044 struct vvp_io vs_ios
;
1047 static inline struct vvp_session
*vvp_env_session(const struct lu_env
*env
)
1049 extern struct lu_context_key vvp_session_key
;
1050 struct vvp_session
*ses
;
1052 ses
= lu_context_key_get(env
->le_ses
, &vvp_session_key
);
1053 LASSERT(ses
!= NULL
);
1057 static inline struct vvp_io
*vvp_env_io(const struct lu_env
*env
)
1059 return &vvp_env_session(env
)->vs_ios
;
1062 void ll_queue_done_writing(struct inode
*inode
, unsigned long flags
);
1063 void ll_close_thread_shutdown(struct ll_close_queue
*lcq
);
1064 int ll_close_thread_start(struct ll_close_queue
**lcq_ret
);
1066 /* llite/llite_mmap.c */
1067 typedef struct rb_root rb_root_t
;
1068 typedef struct rb_node rb_node_t
;
1070 struct ll_lock_tree_node
;
1071 struct ll_lock_tree
{
1073 struct list_head lt_locked_list
;
1074 struct ll_file_data
*lt_fd
;
1077 int ll_teardown_mmaps(struct address_space
*mapping
, __u64 first
, __u64 last
);
1078 int ll_file_mmap(struct file
* file
, struct vm_area_struct
* vma
);
1079 struct ll_lock_tree_node
* ll_node_from_inode(struct inode
*inode
, __u64 start
,
1080 __u64 end
, ldlm_mode_t mode
);
1081 void policy_from_vma(ldlm_policy_data_t
*policy
,
1082 struct vm_area_struct
*vma
, unsigned long addr
, size_t count
);
1083 struct vm_area_struct
*our_vma(struct mm_struct
*mm
, unsigned long addr
,
1086 static inline void ll_invalidate_page(struct page
*vmpage
)
1088 struct address_space
*mapping
= vmpage
->mapping
;
1089 loff_t offset
= vmpage
->index
<< PAGE_CACHE_SHIFT
;
1091 LASSERT(PageLocked(vmpage
));
1092 if (mapping
== NULL
)
1095 ll_teardown_mmaps(mapping
, offset
, offset
+ PAGE_CACHE_SIZE
);
1096 truncate_complete_page(mapping
, vmpage
);
1099 #define ll_s2sbi(sb) (s2lsi(sb)->lsi_llsbi)
1101 /* don't need an addref as the sb_info should be holding one */
1102 static inline struct obd_export
*ll_s2dtexp(struct super_block
*sb
)
1104 return ll_s2sbi(sb
)->ll_dt_exp
;
1107 /* don't need an addref as the sb_info should be holding one */
1108 static inline struct obd_export
*ll_s2mdexp(struct super_block
*sb
)
1110 return ll_s2sbi(sb
)->ll_md_exp
;
1113 static inline struct client_obd
*sbi2mdc(struct ll_sb_info
*sbi
)
1115 struct obd_device
*obd
= sbi
->ll_md_exp
->exp_obd
;
1121 // FIXME: replace the name of this with LL_SB to conform to kernel stuff
1122 static inline struct ll_sb_info
*ll_i2sbi(struct inode
*inode
)
1124 return ll_s2sbi(inode
->i_sb
);
1127 static inline struct obd_export
*ll_i2dtexp(struct inode
*inode
)
1129 return ll_s2dtexp(inode
->i_sb
);
1132 static inline struct obd_export
*ll_i2mdexp(struct inode
*inode
)
1134 return ll_s2mdexp(inode
->i_sb
);
1137 static inline struct lu_fid
*ll_inode2fid(struct inode
*inode
)
1141 LASSERT(inode
!= NULL
);
1142 fid
= &ll_i2info(inode
)->lli_fid
;
1147 static inline int ll_mds_max_easize(struct super_block
*sb
)
1149 return sbi2mdc(ll_s2sbi(sb
))->cl_max_mds_easize
;
1152 static inline __u64
ll_file_maxbytes(struct inode
*inode
)
1154 return ll_i2info(inode
)->lli_maxbytes
;
1158 int ll_setxattr(struct dentry
*dentry
, const char *name
,
1159 const void *value
, size_t size
, int flags
);
1160 ssize_t
ll_getxattr(struct dentry
*dentry
, const char *name
,
1161 void *buffer
, size_t size
);
1162 ssize_t
ll_listxattr(struct dentry
*dentry
, char *buffer
, size_t size
);
1163 int ll_removexattr(struct dentry
*dentry
, const char *name
);
1165 /* llite/remote_perm.c */
1166 extern struct kmem_cache
*ll_remote_perm_cachep
;
1167 extern struct kmem_cache
*ll_rmtperm_hash_cachep
;
1169 struct hlist_head
*alloc_rmtperm_hash(void);
1170 void free_rmtperm_hash(struct hlist_head
*hash
);
1171 int ll_update_remote_perm(struct inode
*inode
, struct mdt_remote_perm
*perm
);
1172 int lustre_check_remote_perm(struct inode
*inode
, int mask
);
1174 /* llite/llite_capa.c */
1175 extern struct timer_list ll_capa_timer
;
1177 int ll_capa_thread_start(void);
1178 void ll_capa_thread_stop(void);
1179 void ll_capa_timer_callback(unsigned long unused
);
1181 struct obd_capa
*ll_add_capa(struct inode
*inode
, struct obd_capa
*ocapa
);
1182 int ll_update_capa(struct obd_capa
*ocapa
, struct lustre_capa
*capa
);
1184 void ll_capa_open(struct inode
*inode
);
1185 void ll_capa_close(struct inode
*inode
);
1187 struct obd_capa
*ll_mdscapa_get(struct inode
*inode
);
1188 struct obd_capa
*ll_osscapa_get(struct inode
*inode
, __u64 opc
);
1190 void ll_truncate_free_capa(struct obd_capa
*ocapa
);
1191 void ll_clear_inode_capas(struct inode
*inode
);
1192 void ll_print_capa_stat(struct ll_sb_info
*sbi
);
1194 /* llite/llite_cl.c */
1195 extern struct lu_device_type vvp_device_type
;
1198 * Common IO arguments for various VFS I/O interfaces.
1200 int cl_sb_init(struct super_block
*sb
);
1201 int cl_sb_fini(struct super_block
*sb
);
1202 enum cl_lock_mode
vvp_mode_from_vma(struct vm_area_struct
*vma
);
1203 void ll_io_init(struct cl_io
*io
, const struct file
*file
, int write
);
1205 void ras_update(struct ll_sb_info
*sbi
, struct inode
*inode
,
1206 struct ll_readahead_state
*ras
, unsigned long index
,
1208 void ll_ra_count_put(struct ll_sb_info
*sbi
, unsigned long len
);
1209 int ll_is_file_contended(struct file
*file
);
1210 void ll_ra_stats_inc(struct address_space
*mapping
, enum ra_stat which
);
1212 /* llite/llite_rmtacl.c */
1213 #ifdef CONFIG_FS_POSIX_ACL
1215 struct list_head ee_list
;
1216 pid_t ee_key
; /* hash key */
1217 struct lu_fid ee_fid
;
1218 int ee_type
; /* ACL type for ACCESS or DEFAULT */
1219 ext_acl_xattr_header
*ee_acl
;
1222 obd_valid
rce_ops2valid(int ops
);
1223 struct rmtacl_ctl_entry
*rct_search(struct rmtacl_ctl_table
*rct
, pid_t key
);
1224 int rct_add(struct rmtacl_ctl_table
*rct
, pid_t key
, int ops
);
1225 int rct_del(struct rmtacl_ctl_table
*rct
, pid_t key
);
1226 void rct_init(struct rmtacl_ctl_table
*rct
);
1227 void rct_fini(struct rmtacl_ctl_table
*rct
);
1229 void ee_free(struct eacl_entry
*ee
);
1230 int ee_add(struct eacl_table
*et
, pid_t key
, struct lu_fid
*fid
, int type
,
1231 ext_acl_xattr_header
*header
);
1232 struct eacl_entry
*et_search_del(struct eacl_table
*et
, pid_t key
,
1233 struct lu_fid
*fid
, int type
);
1234 void et_search_free(struct eacl_table
*et
, pid_t key
);
1235 void et_init(struct eacl_table
*et
);
1236 void et_fini(struct eacl_table
*et
);
1238 static inline obd_valid
rce_ops2valid(int ops
)
1246 #define LL_SA_RPC_MIN 2
1247 #define LL_SA_RPC_DEF 32
1248 #define LL_SA_RPC_MAX 8192
1250 #define LL_SA_CACHE_BIT 5
1251 #define LL_SA_CACHE_SIZE (1 << LL_SA_CACHE_BIT)
1252 #define LL_SA_CACHE_MASK (LL_SA_CACHE_SIZE - 1)
1254 /* per inode struct, for dir only */
1255 struct ll_statahead_info
{
1256 struct inode
*sai_inode
;
1257 atomic_t sai_refcount
; /* when access this struct, hold
1259 unsigned int sai_generation
; /* generation for statahead */
1260 unsigned int sai_max
; /* max ahead of lookup */
1261 __u64 sai_sent
; /* stat requests sent count */
1262 __u64 sai_replied
; /* stat requests which received
1264 __u64 sai_index
; /* index of statahead entry */
1265 __u64 sai_index_wait
; /* index of entry which is the
1266 * caller is waiting for */
1267 __u64 sai_hit
; /* hit count */
1268 __u64 sai_miss
; /* miss count:
1269 * for "ls -al" case, it includes
1270 * hidden dentry miss;
1271 * for "ls -l" case, it does not
1272 * include hidden dentry miss.
1273 * "sai_miss_hidden" is used for
1276 unsigned int sai_consecutive_miss
; /* consecutive miss */
1277 unsigned int sai_miss_hidden
;/* "ls -al", but first dentry
1278 * is not a hidden one */
1279 unsigned int sai_skip_hidden
;/* skipped hidden dentry count */
1280 unsigned int sai_ls_all
:1, /* "ls -al", do stat-ahead for
1282 sai_in_readpage
:1,/* statahead is in readdir()*/
1283 sai_agl_valid
:1;/* AGL is valid for the dir */
1284 wait_queue_head_t sai_waitq
; /* stat-ahead wait queue */
1285 struct ptlrpc_thread sai_thread
; /* stat-ahead thread */
1286 struct ptlrpc_thread sai_agl_thread
; /* AGL thread */
1287 struct list_head sai_entries
; /* entry list */
1288 struct list_head sai_entries_received
; /* entries returned */
1289 struct list_head sai_entries_stated
; /* entries stated */
1290 struct list_head sai_entries_agl
; /* AGL entries to be sent */
1291 struct list_head sai_cache
[LL_SA_CACHE_SIZE
];
1292 spinlock_t sai_cache_lock
[LL_SA_CACHE_SIZE
];
1293 atomic_t sai_cache_count
; /* entry count in cache */
1296 int do_statahead_enter(struct inode
*dir
, struct dentry
**dentry
,
1298 void ll_stop_statahead(struct inode
*dir
, void *key
);
1300 static inline int ll_glimpse_size(struct inode
*inode
)
1302 struct ll_inode_info
*lli
= ll_i2info(inode
);
1305 down_read(&lli
->lli_glimpse_sem
);
1306 rc
= cl_glimpse_size(inode
);
1307 lli
->lli_glimpse_time
= cfs_time_current();
1308 up_read(&lli
->lli_glimpse_sem
);
1313 ll_statahead_mark(struct inode
*dir
, struct dentry
*dentry
)
1315 struct ll_inode_info
*lli
= ll_i2info(dir
);
1316 struct ll_statahead_info
*sai
= lli
->lli_sai
;
1317 struct ll_dentry_data
*ldd
= ll_d2d(dentry
);
1319 /* not the same process, don't mark */
1320 if (lli
->lli_opendir_pid
!= current_pid())
1323 LASSERT(ldd
!= NULL
);
1325 ldd
->lld_sa_generation
= sai
->sai_generation
;
1329 ll_need_statahead(struct inode
*dir
, struct dentry
*dentryp
)
1331 struct ll_inode_info
*lli
;
1332 struct ll_dentry_data
*ldd
;
1334 if (ll_i2sbi(dir
)->ll_sa_max
== 0)
1337 lli
= ll_i2info(dir
);
1338 /* not the same process, don't statahead */
1339 if (lli
->lli_opendir_pid
!= current_pid())
1342 /* statahead has been stopped */
1343 if (lli
->lli_opendir_key
== NULL
)
1346 ldd
= ll_d2d(dentryp
);
1348 * When stats a dentry, the system trigger more than once "revalidate"
1349 * or "lookup", for "getattr", for "getxattr", and maybe for others.
1350 * Under patchless client mode, the operation intent is not accurate,
1351 * which maybe misguide the statahead thread. For example:
1352 * The "revalidate" call for "getattr" and "getxattr" of a dentry maybe
1353 * have the same operation intent -- "IT_GETATTR".
1354 * In fact, one dentry should has only one chance to interact with the
1355 * statahead thread, otherwise the statahead windows will be confused.
1356 * The solution is as following:
1357 * Assign "lld_sa_generation" with "sai_generation" when a dentry
1358 * "IT_GETATTR" for the first time, and the subsequent "IT_GETATTR"
1359 * will bypass interacting with statahead thread for checking:
1360 * "lld_sa_generation == lli_sai->sai_generation"
1362 if (ldd
&& lli
->lli_sai
&&
1363 ldd
->lld_sa_generation
== lli
->lli_sai
->sai_generation
)
1370 ll_statahead_enter(struct inode
*dir
, struct dentry
**dentryp
, int only_unplug
)
1374 ret
= ll_need_statahead(dir
, *dentryp
);
1378 return do_statahead_enter(dir
, dentryp
, only_unplug
);
1381 /* llite ioctl register support rountine */
1387 #define LLIOC_MAX_CMD 256
1390 * Rules to write a callback function:
1393 * @magic: Dynamic ioctl call routine will feed this vaule with the pointer
1394 * returned to ll_iocontrol_register. Callback functions should use this
1395 * data to check the potential collasion of ioctl cmd. If collasion is
1396 * found, callback function should return LLIOC_CONT.
1397 * @rcp: The result of ioctl command.
1400 * If @magic matches the pointer returned by ll_iocontrol_data, the
1401 * callback should return LLIOC_STOP; return LLIOC_STOP otherwise.
1403 typedef enum llioc_iter (*llioc_callback_t
)(struct inode
*inode
,
1404 struct file
*file
, unsigned int cmd
, unsigned long arg
,
1405 void *magic
, int *rcp
);
1407 enum llioc_iter
ll_iocontrol_call(struct inode
*inode
, struct file
*file
,
1408 unsigned int cmd
, unsigned long arg
, int *rcp
);
1410 /* export functions */
1411 /* Register ioctl block dynamatically for a regular file.
1413 * @cmd: the array of ioctl command set
1414 * @count: number of commands in the @cmd
1415 * @cb: callback function, it will be called if an ioctl command is found to
1416 * belong to the command list @cmd.
1419 * A magic pointer will be returned if success;
1420 * otherwise, NULL will be returned.
1422 void *ll_iocontrol_register(llioc_callback_t cb
, int count
, unsigned int *cmd
);
1423 void ll_iocontrol_unregister(void *magic
);
1426 /* lclient compat stuff */
1427 #define cl_inode_info ll_inode_info
1428 #define cl_i2info(info) ll_i2info(info)
1429 #define cl_inode_mode(inode) ((inode)->i_mode)
1430 #define cl_i2sbi ll_i2sbi
1432 static inline struct ll_file_data
*cl_iattr2fd(struct inode
*inode
,
1433 const struct iattr
*attr
)
1435 LASSERT(attr
->ia_valid
& ATTR_FILE
);
1436 return LUSTRE_FPRIVATE(attr
->ia_file
);
1439 static inline void cl_isize_lock(struct inode
*inode
)
1441 ll_inode_size_lock(inode
);
1444 static inline void cl_isize_unlock(struct inode
*inode
)
1446 ll_inode_size_unlock(inode
);
1449 static inline void cl_isize_write_nolock(struct inode
*inode
, loff_t kms
)
1451 LASSERT(down_trylock(&ll_i2info(inode
)->lli_size_sem
) != 0);
1452 i_size_write(inode
, kms
);
1455 static inline void cl_isize_write(struct inode
*inode
, loff_t kms
)
1457 ll_inode_size_lock(inode
);
1458 i_size_write(inode
, kms
);
1459 ll_inode_size_unlock(inode
);
1462 #define cl_isize_read(inode) i_size_read(inode)
1464 static inline int cl_merge_lvb(const struct lu_env
*env
, struct inode
*inode
)
1466 return ll_merge_lvb(env
, inode
);
1469 #define cl_inode_atime(inode) LTIME_S((inode)->i_atime)
1470 #define cl_inode_ctime(inode) LTIME_S((inode)->i_ctime)
1471 #define cl_inode_mtime(inode) LTIME_S((inode)->i_mtime)
1473 struct obd_capa
*cl_capa_lookup(struct inode
*inode
, enum cl_req_type crt
);
1475 int cl_sync_file_range(struct inode
*inode
, loff_t start
, loff_t end
,
1476 enum cl_fsync_mode mode
, int ignore_layout
);
1478 /** direct write pages */
1479 struct ll_dio_pages
{
1480 /** page array to be written. we don't support
1481 * partial pages except the last one. */
1482 struct page
**ldp_pages
;
1483 /* offset of each page */
1484 loff_t
*ldp_offsets
;
1485 /** if ldp_offsets is NULL, it means a sequential
1486 * pages to be written, then this is the file offset
1487 * of the * first page. */
1488 loff_t ldp_start_offset
;
1489 /** how many bytes are to be written. */
1491 /** # of pages in the array. */
1495 static inline void cl_stats_tally(struct cl_device
*dev
, enum cl_req_type crt
,
1498 int opc
= (crt
== CRT_READ
) ? LPROC_LL_OSC_READ
:
1501 ll_stats_ops_tally(ll_s2sbi(cl2ccc_dev(dev
)->cdv_sb
), opc
, rc
);
1504 extern ssize_t
ll_direct_rw_pages(const struct lu_env
*env
, struct cl_io
*io
,
1505 int rw
, struct inode
*inode
,
1506 struct ll_dio_pages
*pv
);
1508 static inline int ll_file_nolock(const struct file
*file
)
1510 struct ll_file_data
*fd
= LUSTRE_FPRIVATE(file
);
1511 struct inode
*inode
= file
->f_dentry
->d_inode
;
1513 LASSERT(fd
!= NULL
);
1514 return ((fd
->fd_flags
& LL_FILE_IGNORE_LOCK
) ||
1515 (ll_i2sbi(inode
)->ll_flags
& LL_SBI_NOLCK
));
1518 static inline void ll_set_lock_data(struct obd_export
*exp
, struct inode
*inode
,
1519 struct lookup_intent
*it
, __u64
*bits
)
1521 if (!it
->d
.lustre
.it_lock_set
) {
1522 struct lustre_handle handle
;
1524 /* If this inode is a remote object, it will get two
1525 * separate locks in different namespaces, Master MDT,
1526 * where the name entry is, will grant LOOKUP lock,
1527 * remote MDT, where the object is, will grant
1528 * UPDATE|PERM lock. The inode will be attched to both
1529 * LOOKUP and PERM locks, so revoking either locks will
1530 * case the dcache being cleared */
1531 if (it
->d
.lustre
.it_remote_lock_mode
) {
1532 handle
.cookie
= it
->d
.lustre
.it_remote_lock_handle
;
1533 CDEBUG(D_DLMTRACE
, "setting l_data to inode %p"
1534 "(%lu/%u) for remote lock "LPX64
"\n", inode
,
1535 inode
->i_ino
, inode
->i_generation
,
1537 md_set_lock_data(exp
, &handle
.cookie
, inode
, NULL
);
1540 handle
.cookie
= it
->d
.lustre
.it_lock_handle
;
1542 CDEBUG(D_DLMTRACE
, "setting l_data to inode %p (%lu/%u)"
1543 " for lock "LPX64
"\n", inode
, inode
->i_ino
,
1544 inode
->i_generation
, handle
.cookie
);
1546 md_set_lock_data(exp
, &handle
.cookie
, inode
,
1547 &it
->d
.lustre
.it_lock_bits
);
1548 it
->d
.lustre
.it_lock_set
= 1;
1552 *bits
= it
->d
.lustre
.it_lock_bits
;
1555 static inline void ll_lock_dcache(struct inode
*inode
)
1557 spin_lock(&inode
->i_lock
);
1560 static inline void ll_unlock_dcache(struct inode
*inode
)
1562 spin_unlock(&inode
->i_lock
);
1565 static inline int d_lustre_invalid(const struct dentry
*dentry
)
1567 struct ll_dentry_data
*lld
= ll_d2d(dentry
);
1569 return (lld
== NULL
) || lld
->lld_invalid
;
1572 static inline void __d_lustre_invalidate(struct dentry
*dentry
)
1574 struct ll_dentry_data
*lld
= ll_d2d(dentry
);
1577 lld
->lld_invalid
= 1;
1581 * Mark dentry INVALID, if dentry refcount is zero (this is normally case for
1582 * ll_md_blocking_ast), unhash this dentry, and let dcache to reclaim it later;
1583 * else dput() of the last refcount will unhash this dentry and kill it.
1585 static inline void d_lustre_invalidate(struct dentry
*dentry
, int nested
)
1587 CDEBUG(D_DENTRY
, "invalidate dentry %.*s (%p) parent %p inode %p "
1588 "refc %d\n", dentry
->d_name
.len
, dentry
->d_name
.name
, dentry
,
1589 dentry
->d_parent
, dentry
->d_inode
, d_count(dentry
));
1591 spin_lock_nested(&dentry
->d_lock
,
1592 nested
? DENTRY_D_LOCK_NESTED
: DENTRY_D_LOCK_NORMAL
);
1593 __d_lustre_invalidate(dentry
);
1594 if (d_count(dentry
) == 0)
1596 spin_unlock(&dentry
->d_lock
);
1599 static inline void d_lustre_revalidate(struct dentry
*dentry
)
1601 spin_lock(&dentry
->d_lock
);
1602 LASSERT(ll_d2d(dentry
) != NULL
);
1603 ll_d2d(dentry
)->lld_invalid
= 0;
1604 spin_unlock(&dentry
->d_lock
);
1607 #if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 7, 50, 0)
1608 /* Compatibility for old (1.8) compiled userspace quota code */
1609 struct if_quotactl_18
{
1614 struct obd_dqinfo qc_dqinfo
;
1615 struct obd_dqblk qc_dqblk
;
1617 struct obd_uuid obd_uuid
;
1619 #define LL_IOC_QUOTACTL_18 _IOWR('f', 162, struct if_quotactl_18 *)
1620 /* End compatibility for old (1.8) compiled userspace quota code */
1622 #warning "remove old LL_IOC_QUOTACTL_18 compatibility code"
1623 #endif /* LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 7, 50, 0) */
1626 LL_LAYOUT_GEN_NONE
= ((__u32
)-2), /* layout lock was cancelled */
1627 LL_LAYOUT_GEN_EMPTY
= ((__u32
)-1) /* for empty layout */
1630 int ll_layout_conf(struct inode
*inode
, const struct cl_object_conf
*conf
);
1631 int ll_layout_refresh(struct inode
*inode
, __u32
*gen
);
1632 int ll_layout_restore(struct inode
*inode
);
1634 int ll_xattr_init(void);
1635 void ll_xattr_fini(void);
1637 #endif /* LLITE_INTERNAL_H */