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.gnu.org/licenses/gpl-2.0.html
23 * Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved.
24 * Use is subject to license terms.
26 * Copyright (c) 2011, 2017, Intel Corporation.
29 * This file is part of Lustre, http://www.lustre.org/
30 * Lustre is a trademark of Sun Microsystems, Inc.
33 #ifndef LLITE_INTERNAL_H
34 #define LLITE_INTERNAL_H
35 #include <lustre_debug.h>
36 #include <lustre_disk.h> /* for s2sbi */
37 #include <lustre_eacl.h>
38 #include <lustre_linkea.h>
40 /* for struct cl_lock_descr and struct cl_io */
41 #include <cl_object.h>
42 #include <lustre_lmv.h>
43 #include <lustre_mdc.h>
44 #include <lustre_intent.h>
45 #include <linux/compat.h>
46 #include <linux/aio.h>
47 #include <lustre_compat.h>
49 #include "vvp_internal.h"
50 #include "range_lock.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 /* 4UL * 1024 * 1024 */
72 #define LL_MAX_BLKSIZE_BITS 22
74 #define LL_IT2STR(it) ((it) ? ldlm_it2str((it)->it_op) : "0")
75 #define LUSTRE_FPRIVATE(file) ((file)->private_data)
77 struct ll_dentry_data {
78 struct lookup_intent *lld_it;
79 unsigned int lld_sa_generation;
80 unsigned int lld_invalid:1;
81 unsigned int lld_nfs_dentry:1;
82 struct rcu_head lld_rcu_head;
85 #define ll_d2d(de) ((struct ll_dentry_data*)((de)->d_fsdata))
87 #define LLI_INODE_MAGIC 0x111d0de5
88 #define LLI_INODE_DEAD 0xdeadd00d
90 struct ll_getname_data {
91 #ifdef HAVE_DIR_CONTEXT
92 struct dir_context ctx;
94 char *lgd_name; /* points to a buffer with NAME_MAX+1 size */
95 struct lu_fid lgd_fid; /* target fid we are looking for */
96 int lgd_found; /* inode matched? */
100 struct lu_env *lg_env;
102 struct cl_lock *lg_lock;
103 unsigned long lg_gid;
106 struct ll_inode_info {
107 __u32 lli_inode_magic;
110 volatile unsigned long lli_flags;
111 struct posix_acl *lli_posix_acl;
113 /* identifying fields for both metadata and data stacks. */
114 struct lu_fid lli_fid;
115 /* master inode fid for stripe directory */
116 struct lu_fid lli_pfid;
118 /* We need all three because every inode may be opened in different
120 struct obd_client_handle *lli_mds_read_och;
121 struct obd_client_handle *lli_mds_write_och;
122 struct obd_client_handle *lli_mds_exec_och;
123 __u64 lli_open_fd_read_count;
124 __u64 lli_open_fd_write_count;
125 __u64 lli_open_fd_exec_count;
126 /* Protects access to och pointers and their usage counters */
127 struct mutex lli_och_mutex;
129 struct inode lli_vfs_inode;
131 /* the most recent timestamps obtained from mds */
135 spinlock_t lli_agl_lock;
137 /* update atime from MDS no matter if it's older than
138 * local inode atime. */
139 unsigned int lli_update_atime:1;
141 /* Try to make the d::member and f::member are aligned. Before using
142 * these members, make clear whether it is directory or not. */
146 /* metadata statahead */
147 /* since parent-child threads can share the same @file
148 * struct, "opendir_key" is the token when dir close for
149 * case of parent exit before child -- it is me should
150 * cleanup the dir readahead. */
151 void *lli_opendir_key;
152 struct ll_statahead_info *lli_sai;
153 /* protect statahead stuff. */
154 spinlock_t lli_sa_lock;
155 /* "opendir_pid" is the token when lookup/revalid
156 * -- I am the owner of dir statahead. */
157 pid_t lli_opendir_pid;
158 /* stat will try to access statahead entries or start
159 * statahead if this flag is set, and this flag will be
160 * set upon dir open, and cleared when dir is closed,
161 * statahead hit ratio is too low, or start statahead
163 unsigned int lli_sa_enabled:1;
164 /* generation for statahead */
165 unsigned int lli_sa_generation;
166 /* rw lock protects lli_lsm_md */
167 struct rw_semaphore lli_lsm_sem;
168 /* directory stripe information */
169 struct lmv_stripe_md *lli_lsm_md;
170 /* directory default LMV */
171 struct lmv_stripe_md *lli_default_lsm_md;
174 /* for non-directory */
176 struct mutex lli_size_mutex;
177 char *lli_symlink_name;
179 * struct rw_semaphore {
180 * signed long count; // align d.d_def_acl
181 * spinlock_t wait_lock; // align d.d_sa_lock
182 * struct list_head wait_list;
185 struct rw_semaphore lli_trunc_sem;
186 struct range_lock_tree lli_write_tree;
188 struct rw_semaphore lli_glimpse_sem;
189 ktime_t lli_glimpse_time;
190 struct list_head lli_agl_list;
193 /* for writepage() only to communicate to fsync */
196 /* protect the file heat fields */
197 spinlock_t lli_heat_lock;
198 __u32 lli_heat_flags;
199 struct obd_heat_instance lli_heat_instances[OBD_HEAT_COUNT];
202 * Whenever a process try to read/write the file, the
203 * jobid of the process will be saved here, and it'll
204 * be packed into the write PRC when flush later.
206 * So the read/write statistics for jobid will not be
207 * accurate if the file is shared by different jobs.
209 char lli_jobid[LUSTRE_JOBID_SIZE];
211 struct mutex lli_pcc_lock;
212 enum lu_pcc_state_flags lli_pcc_state;
213 struct pcc_inode *lli_pcc_inode;
214 struct mutex lli_group_mutex;
215 __u64 lli_group_users;
216 unsigned long lli_group_gid;
220 /* XXX: For following frequent used members, although they maybe special
221 * used for non-directory object, it is some time-wasting to check
222 * whether the object is directory or not before using them. On the
223 * other hand, currently, sizeof(f) > sizeof(d), it cannot reduce
224 * the "ll_inode_info" size even if moving those members into u.f.
225 * So keep them out side.
227 * In the future, if more members are added only for directory,
228 * some of the following members can be moved into u.f.
230 struct cl_object *lli_clob;
232 /* mutex to request for layout lock exclusively. */
233 struct mutex lli_layout_mutex;
234 /* Layout version, protected by lli_layout_lock */
235 __u32 lli_layout_gen;
236 spinlock_t lli_layout_lock;
238 __u32 lli_projid; /* project id */
240 struct rw_semaphore lli_xattrs_list_rwsem;
241 struct mutex lli_xattrs_enq_lock;
242 struct list_head lli_xattrs; /* ll_xattr_entry->xe_list */
245 static inline __u32 ll_layout_version_get(struct ll_inode_info *lli)
249 spin_lock(&lli->lli_layout_lock);
250 gen = lli->lli_layout_gen;
251 spin_unlock(&lli->lli_layout_lock);
256 static inline void ll_layout_version_set(struct ll_inode_info *lli, __u32 gen)
258 spin_lock(&lli->lli_layout_lock);
259 lli->lli_layout_gen = gen;
260 spin_unlock(&lli->lli_layout_lock);
264 /* File data is modified. */
265 LLIF_DATA_MODIFIED = 0,
266 /* File is being restored */
267 LLIF_FILE_RESTORING = 1,
268 /* Xattr cache is attached to the file */
269 LLIF_XATTR_CACHE = 2,
270 /* Project inherit */
271 LLIF_PROJECT_INHERIT = 3,
274 static inline void ll_file_set_flag(struct ll_inode_info *lli,
275 enum ll_file_flags flag)
277 set_bit(flag, &lli->lli_flags);
280 static inline void ll_file_clear_flag(struct ll_inode_info *lli,
281 enum ll_file_flags flag)
283 clear_bit(flag, &lli->lli_flags);
286 static inline bool ll_file_test_flag(struct ll_inode_info *lli,
287 enum ll_file_flags flag)
289 return test_bit(flag, &lli->lli_flags);
292 static inline bool ll_file_test_and_clear_flag(struct ll_inode_info *lli,
293 enum ll_file_flags flag)
295 return test_and_clear_bit(flag, &lli->lli_flags);
298 int ll_xattr_cache_destroy(struct inode *inode);
300 int ll_xattr_cache_get(struct inode *inode,
306 static inline bool obd_connect_has_secctx(struct obd_connect_data *data)
308 #if defined(HAVE_SECURITY_DENTRY_INIT_SECURITY) && defined(CONFIG_SECURITY)
309 return data->ocd_connect_flags & OBD_CONNECT_FLAGS2 &&
310 data->ocd_connect_flags2 & OBD_CONNECT2_FILE_SECCTX;
313 #endif /* HAVE_SECURITY_DENTRY_INIT_SECURITY */
316 static inline void obd_connect_set_secctx(struct obd_connect_data *data)
318 #if defined(HAVE_SECURITY_DENTRY_INIT_SECURITY) && defined(CONFIG_SECURITY)
319 data->ocd_connect_flags2 |= OBD_CONNECT2_FILE_SECCTX;
320 #endif /* HAVE_SECURITY_DENTRY_INIT_SECURITY */
323 int ll_dentry_init_security(struct dentry *dentry, int mode, struct qstr *name,
324 const char **secctx_name, void **secctx,
326 int ll_inode_init_security(struct dentry *dentry, struct inode *inode,
329 int ll_listsecurity(struct inode *inode, char *secctx_name,
330 size_t secctx_name_size);
333 * Locking to guarantee consistency of non-atomic updates to long long i_size,
334 * consistency between file size and KMS.
336 * Implemented by ->lli_size_mutex and ->lsm_lock, nested in that order.
339 void ll_inode_size_lock(struct inode *inode);
340 void ll_inode_size_unlock(struct inode *inode);
342 static inline struct ll_inode_info *ll_i2info(struct inode *inode)
344 return container_of(inode, struct ll_inode_info, lli_vfs_inode);
347 static inline struct pcc_inode *ll_i2pcci(struct inode *inode)
349 return ll_i2info(inode)->lli_pcc_inode;
352 /* default to use at least 16M for fast read if possible */
353 #define RA_REMAIN_WINDOW_MIN MiB_TO_PAGES(16UL)
355 /* default to about 64M of readahead on a given system. */
356 #define SBI_DEFAULT_READAHEAD_MAX MiB_TO_PAGES(64UL)
358 /* default to read-ahead full files smaller than 2MB on the second read */
359 #define SBI_DEFAULT_READAHEAD_WHOLE_MAX MiB_TO_PAGES(2UL)
364 RA_STAT_DISTANT_READPAGE,
365 RA_STAT_MISS_IN_WINDOW,
366 RA_STAT_FAILED_GRAB_PAGE,
367 RA_STAT_FAILED_MATCH,
372 RA_STAT_MAX_IN_FLIGHT,
373 RA_STAT_WRONG_GRAB_PAGE,
374 RA_STAT_FAILED_REACH_END,
376 RA_STAT_FAILED_FAST_READ,
381 atomic_t ra_cur_pages;
382 unsigned long ra_max_pages;
383 unsigned long ra_max_pages_per_file;
384 unsigned long ra_max_read_ahead_whole_pages;
385 struct workqueue_struct *ll_readahead_wq;
387 * Max number of active works for readahead workqueue,
388 * default is 0 which make workqueue init number itself,
389 * unless there is a specific need for throttling the
390 * number of active work items, specifying '0' is recommended.
392 unsigned int ra_async_max_active;
393 /* Threshold to control when to trigger async readahead */
394 unsigned long ra_async_pages_per_file_threshold;
397 /* ra_io_arg will be filled in the beginning of ll_readahead with
398 * ras_lock, then the following ll_read_ahead_pages will read RA
399 * pages according to this arg, all the items in this structure are
400 * counted by page index.
403 pgoff_t ria_start; /* start offset of read-ahead*/
404 pgoff_t ria_end; /* end offset of read-ahead*/
405 unsigned long ria_reserved; /* reserved pages for read-ahead */
406 pgoff_t ria_end_min; /* minimum end to cover current read */
407 bool ria_eof; /* reach end of file */
408 /* If stride read pattern is detected, ria_stoff means where
409 * stride read is started. Note: for normal read-ahead, the
410 * value here is meaningless, and also it will not be accessed*/
411 unsigned long ria_stoff;
412 /* ria_length and ria_bytes are the length and pages length in the
413 * stride I/O mode. And they will also be used to check whether
414 * it is stride I/O read-ahead in the read-ahead pages*/
415 unsigned long ria_length;
416 unsigned long ria_bytes;
419 /* LL_HIST_MAX=32 causes an overflow */
420 #define LL_HIST_MAX 28
421 #define LL_HIST_START 12 /* buckets start at 2^12 = 4k */
422 #define LL_PROCESS_HIST_MAX 10
423 struct per_process_info {
425 struct obd_histogram pp_r_hist;
426 struct obd_histogram pp_w_hist;
429 /* pp_extents[LL_PROCESS_HIST_MAX] will hold the combined process info */
430 struct ll_rw_extents_info {
431 struct per_process_info pp_extents[LL_PROCESS_HIST_MAX + 1];
434 #define LL_OFFSET_HIST_MAX 100
435 struct ll_rw_process_info {
438 loff_t rw_range_start;
440 loff_t rw_last_file_pos;
442 size_t rw_smallest_extent;
443 size_t rw_largest_extent;
444 struct ll_file_data *rw_last_file;
447 enum stats_track_type {
448 STATS_TRACK_ALL = 0, /* track all processes */
449 STATS_TRACK_PID, /* track process with this pid */
450 STATS_TRACK_PPID, /* track processes with this ppid */
451 STATS_TRACK_GID, /* track processes with this gid */
455 /* flags for sbi->ll_flags */
456 #define LL_SBI_NOLCK 0x01 /* DLM locking disabled (directio-only) */
457 #define LL_SBI_CHECKSUM 0x02 /* checksum each page as it's written */
458 #define LL_SBI_FLOCK 0x04
459 #define LL_SBI_USER_XATTR 0x08 /* support user xattr */
460 #define LL_SBI_ACL 0x10 /* support ACL */
461 /* LL_SBI_RMT_CLIENT 0x40 remote client */
462 #define LL_SBI_MDS_CAPA 0x80 /* support mds capa, obsolete */
463 #define LL_SBI_OSS_CAPA 0x100 /* support oss capa, obsolete */
464 #define LL_SBI_LOCALFLOCK 0x200 /* Local flocks support by kernel */
465 #define LL_SBI_LRU_RESIZE 0x400 /* lru resize support */
466 #define LL_SBI_LAZYSTATFS 0x800 /* lazystatfs mount option */
467 /* LL_SBI_SOM_PREVIEW 0x1000 SOM preview mount option, obsolete */
468 #define LL_SBI_32BIT_API 0x2000 /* generate 32 bit inodes. */
469 #define LL_SBI_64BIT_HASH 0x4000 /* support 64-bits dir hash/offset */
470 #define LL_SBI_AGL_ENABLED 0x8000 /* enable agl */
471 #define LL_SBI_VERBOSE 0x10000 /* verbose mount/umount */
472 #define LL_SBI_LAYOUT_LOCK 0x20000 /* layout lock support */
473 #define LL_SBI_USER_FID2PATH 0x40000 /* allow fid2path by unprivileged users */
474 #define LL_SBI_XATTR_CACHE 0x80000 /* support for xattr cache */
475 #define LL_SBI_NOROOTSQUASH 0x100000 /* do not apply root squash */
476 #define LL_SBI_ALWAYS_PING 0x200000 /* always ping even if server
478 #define LL_SBI_FAST_READ 0x400000 /* fast read support */
479 #define LL_SBI_FILE_SECCTX 0x800000 /* set file security context at create */
480 /* LL_SBI_PIO 0x1000000 parallel IO support, introduced in
482 #define LL_SBI_TINY_WRITE 0x2000000 /* tiny write support */
483 #define LL_SBI_FILE_HEAT 0x4000000 /* file heat support */
484 #define LL_SBI_FLAGS { \
514 /* This is embedded into llite super-blocks to keep track of connect
515 * flags (capabilities) supported by all imports given mount is
517 struct lustre_client_ocd {
518 /* This is conjunction of connect_flags across all imports
519 * (LOVs) this mount is connected to. This field is updated by
520 * cl_ocd_update() under ->lco_lock. */
522 struct mutex lco_lock;
523 struct obd_export *lco_md_exp;
524 struct obd_export *lco_dt_exp;
528 /* this protects pglist and ra_info. It isn't safe to
529 * grab from interrupt contexts */
531 spinlock_t ll_pp_extent_lock; /* pp_extent entry*/
532 spinlock_t ll_process_lock; /* ll_rw_process_info */
533 struct obd_uuid ll_sb_uuid;
534 struct obd_export *ll_md_exp;
535 struct obd_export *ll_dt_exp;
536 struct obd_device *ll_md_obd;
537 struct obd_device *ll_dt_obd;
538 struct dentry *ll_debugfs_entry;
539 struct lu_fid ll_root_fid; /* root object fid */
542 unsigned int ll_xattr_cache_enabled:1,
543 ll_xattr_cache_set:1, /* already set to 0/1 */
544 ll_client_common_fill_super_succeeded:1,
547 struct lustre_client_ocd ll_lco;
549 struct lprocfs_stats *ll_stats; /* lprocfs stats counter */
551 /* Used to track "unstable" pages on a client, and maintain a
552 * LRU list of clean pages. An "unstable" page is defined as
553 * any page which is sent to a server as part of a bulk request,
554 * but is uncommitted to stable storage. */
555 struct cl_client_cache *ll_cache;
557 struct lprocfs_stats *ll_ra_stats;
559 struct ll_ra_info ll_ra_info;
560 unsigned int ll_namelen;
561 struct file_operations *ll_fop;
563 struct lu_site *ll_site;
564 struct cl_device *ll_cl;
566 struct ll_rw_extents_info ll_rw_extents_info;
567 int ll_extent_process_count;
568 struct ll_rw_process_info ll_rw_process_info[LL_PROCESS_HIST_MAX];
569 unsigned int ll_offset_process_count;
570 struct ll_rw_process_info ll_rw_offset_info[LL_OFFSET_HIST_MAX];
571 unsigned int ll_rw_offset_entry_count;
572 int ll_stats_track_id;
573 enum stats_track_type ll_stats_track_type;
576 /* metadata stat-ahead */
577 unsigned int ll_sa_running_max;/* max concurrent
578 * statahead instances */
579 unsigned int ll_sa_max; /* max statahead RPCs */
580 atomic_t ll_sa_total; /* statahead thread started
582 atomic_t ll_sa_wrong; /* statahead thread stopped for
584 atomic_t ll_sa_running; /* running statahead thread
586 atomic_t ll_agl_total; /* AGL thread started count */
588 dev_t ll_sdev_orig; /* save s_dev before assign for
591 struct root_squash_info ll_squash;
594 /* st_blksize returned by stat(2), when non-zero */
595 unsigned int ll_stat_blksize;
597 /* maximum relative age of cached statfs results */
598 unsigned int ll_statfs_max_age;
600 struct kset ll_kset; /* sysfs object */
601 struct completion ll_kobj_unregister;
604 unsigned int ll_heat_decay_weight;
605 unsigned int ll_heat_period_second;
607 /* filesystem fsname */
608 char ll_fsname[LUSTRE_MAXFSNAME + 1];
610 /* Persistent Client Cache */
611 struct pcc_super ll_pcc_super;
614 #define SBI_DEFAULT_HEAT_DECAY_WEIGHT ((80 * 256 + 50) / 100)
615 #define SBI_DEFAULT_HEAT_PERIOD_SECOND (60)
617 * per file-descriptor read-ahead data.
619 struct ll_readahead_state {
621 /* End byte that read(2) try to read. */
622 unsigned long ras_last_read_end;
624 * number of bytes read after last read-ahead window reset. As window
625 * is reset on each seek, this is effectively a number of consecutive
626 * accesses. Maybe ->ras_accessed_in_window is better name.
628 * XXX nikita: window is also reset (by ras_update()) when Lustre
629 * believes that memory pressure evicts read-ahead pages. In that
630 * case, it probably doesn't make sense to expand window to
631 * PTLRPC_MAX_BRW_PAGES on the third access.
633 unsigned long ras_consecutive_bytes;
635 * number of read requests after the last read-ahead window reset
636 * As window is reset on each seek, this is effectively the number
637 * on consecutive read request and is used to trigger read-ahead.
639 unsigned long ras_consecutive_requests;
641 * Parameters of current read-ahead window. Handled by
642 * ras_update(). On the initial access to the file or after a seek,
643 * window is reset to 0. After 3 consecutive accesses, window is
644 * expanded to PTLRPC_MAX_BRW_PAGES. Afterwards, window is enlarged by
645 * PTLRPC_MAX_BRW_PAGES chunks up to ->ra_max_pages.
647 pgoff_t ras_window_start, ras_window_len;
649 * Optimal RPC size. It decides how many pages will be sent
650 * for each read-ahead.
652 unsigned long ras_rpc_size;
654 * Where next read-ahead should start at. This lies within read-ahead
655 * window. Read-ahead window is read in pieces rather than at once
656 * because: 1. lustre limits total number of pages under read-ahead by
657 * ->ra_max_pages (see ll_ra_count_get()), 2. client cannot read pages
658 * not covered by DLM lock.
660 pgoff_t ras_next_readahead;
662 * Total number of ll_file_read requests issued, reads originating
663 * due to mmap are not counted in this total. This value is used to
664 * trigger full file read-ahead after multiple reads to a small file.
666 unsigned long ras_requests;
668 * Page index with respect to the current request, these value
669 * will not be accurate when dealing with reads issued via mmap.
671 unsigned long ras_request_index;
673 * The following 3 items are used for detecting the stride I/O
675 * In stride I/O mode,
676 * ...............|-----data-----|****gap*****|--------|******|....
677 * offset |-stride_bytes-|-stride_gap-|
678 * ras_stride_offset = offset;
679 * ras_stride_length = stride_bytes + stride_gap;
680 * ras_stride_bytes = stride_bytes;
681 * Note: all these three items are counted by bytes.
683 unsigned long ras_stride_length;
684 unsigned long ras_stride_bytes;
685 unsigned long ras_stride_offset;
687 * number of consecutive stride request count, and it is similar as
688 * ras_consecutive_requests, but used for stride I/O mode.
689 * Note: only more than 2 consecutive stride request are detected,
690 * stride read-ahead will be enable
692 unsigned long ras_consecutive_stride_requests;
693 /* index of the last page that async readahead starts */
694 pgoff_t ras_async_last_readpage;
697 struct ll_readahead_work {
698 /** File to readahead */
699 struct file *lrw_file;
701 unsigned long lrw_start;
703 unsigned long lrw_end;
705 /* async worker to handler read */
706 struct work_struct lrw_readahead_work;
709 extern struct kmem_cache *ll_file_data_slab;
710 struct lustre_handle;
711 struct ll_file_data {
712 struct ll_readahead_state fd_ras;
713 struct ll_grouplock fd_grouplock;
717 /* openhandle if lease exists for this file.
718 * Borrow lli->lli_och_mutex to protect assignment */
719 struct obd_client_handle *fd_lease_och;
720 struct obd_client_handle *fd_och;
721 struct file *fd_file;
722 /* Indicate whether need to report failure when close.
723 * true: failure is known, not report again.
724 * false: unknown failure, should report. */
725 bool fd_write_failed;
726 bool ll_lock_no_expand;
727 rwlock_t fd_lock; /* protect lcc list */
728 struct list_head fd_lccs; /* list of ll_cl_context */
729 /* Used by mirrored file to lead IOs to a specific mirror, usually
730 * for mirror resync. 0 means default. */
731 __u32 fd_designated_mirror;
732 /* The layout version when resync starts. Resync I/O should carry this
733 * layout version for verification to OST objects */
734 __u32 fd_layout_version;
735 struct pcc_file fd_pcc_file;
738 void llite_tunables_unregister(void);
739 int llite_tunables_register(void);
741 static inline struct inode *ll_info2i(struct ll_inode_info *lli)
743 return &lli->lli_vfs_inode;
746 __u32 ll_i2suppgid(struct inode *i);
747 void ll_i2gids(__u32 *suppgids, struct inode *i1,struct inode *i2);
749 static inline int ll_need_32bit_api(struct ll_sb_info *sbi)
751 #if BITS_PER_LONG == 32
753 #elif defined(CONFIG_COMPAT)
754 if (unlikely(sbi->ll_flags & LL_SBI_32BIT_API))
757 # ifdef CONFIG_X86_X32
758 /* in_compat_syscall() returns true when called from a kthread
759 * and CONFIG_X86_X32 is enabled, which is wrong. So check
760 * whether the caller comes from a syscall (ie. not a kthread)
761 * before calling in_compat_syscall(). */
762 if (current->flags & PF_KTHREAD)
766 return unlikely(in_compat_syscall());
768 return unlikely(sbi->ll_flags & LL_SBI_32BIT_API);
772 static inline bool ll_sbi_has_fast_read(struct ll_sb_info *sbi)
774 return !!(sbi->ll_flags & LL_SBI_FAST_READ);
777 static inline bool ll_sbi_has_tiny_write(struct ll_sb_info *sbi)
779 return !!(sbi->ll_flags & LL_SBI_TINY_WRITE);
782 static inline bool ll_sbi_has_file_heat(struct ll_sb_info *sbi)
784 return !!(sbi->ll_flags & LL_SBI_FILE_HEAT);
787 void ll_ras_enter(struct file *f);
789 /* llite/lcommon_misc.c */
790 int cl_ocd_update(struct obd_device *host, struct obd_device *watched,
791 enum obd_notify_event ev, void *owner);
792 int cl_get_grouplock(struct cl_object *obj, unsigned long gid, int nonblock,
793 struct ll_grouplock *lg);
794 void cl_put_grouplock(struct ll_grouplock *lg);
796 /* llite/lproc_llite.c */
797 int ll_debugfs_register_super(struct super_block *sb, const char *name);
798 void ll_debugfs_unregister_super(struct super_block *sb);
799 void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count);
803 LPROC_LL_WRITE_BYTES,
826 LPROC_LL_ALLOC_INODE,
829 LPROC_LL_GETXATTR_HITS,
831 LPROC_LL_REMOVEXATTR,
833 LPROC_LL_FILE_OPCODES
837 struct ll_dir_chain {
840 static inline void ll_dir_chain_init(struct ll_dir_chain *chain)
844 static inline void ll_dir_chain_fini(struct ll_dir_chain *chain)
848 extern const struct file_operations ll_dir_operations;
849 extern const struct inode_operations ll_dir_inode_operations;
850 #ifdef HAVE_DIR_CONTEXT
851 int ll_dir_read(struct inode *inode, __u64 *pos, struct md_op_data *op_data,
852 struct dir_context *ctx);
854 int ll_dir_read(struct inode *inode, __u64 *pos, struct md_op_data *op_data,
855 void *cookie, filldir_t filldir);
857 int ll_get_mdt_idx(struct inode *inode);
858 int ll_get_mdt_idx_by_fid(struct ll_sb_info *sbi, const struct lu_fid *fid);
859 struct page *ll_get_dir_page(struct inode *dir, struct md_op_data *op_data,
860 __u64 offset, struct ll_dir_chain *chain);
861 void ll_release_page(struct inode *inode, struct page *page, bool remove);
864 extern const struct inode_operations ll_special_inode_operations;
866 struct inode *ll_iget(struct super_block *sb, ino_t hash,
867 struct lustre_md *lic);
868 int ll_test_inode_by_fid(struct inode *inode, void *opaque);
869 int ll_md_blocking_ast(struct ldlm_lock *, struct ldlm_lock_desc *,
870 void *data, int flag);
871 struct dentry *ll_splice_alias(struct inode *inode, struct dentry *de);
872 int ll_rmdir_entry(struct inode *dir, char *name, int namelen);
873 void ll_update_times(struct ptlrpc_request *request, struct inode *inode);
876 int ll_writepage(struct page *page, struct writeback_control *wbc);
877 int ll_writepages(struct address_space *, struct writeback_control *wbc);
878 int ll_readpage(struct file *file, struct page *page);
879 int ll_io_read_page(const struct lu_env *env, struct cl_io *io,
880 struct cl_page *page, struct file *file);
881 void ll_readahead_init(struct inode *inode, struct ll_readahead_state *ras);
882 int vvp_io_write_commit(const struct lu_env *env, struct cl_io *io);
885 void ll_cl_add(struct file *file, const struct lu_env *env, struct cl_io *io,
887 void ll_cl_remove(struct file *file, const struct lu_env *env);
888 struct ll_cl_context *ll_cl_find(struct file *file);
890 extern const struct address_space_operations ll_aops;
893 extern struct file_operations ll_file_operations;
894 extern struct file_operations ll_file_operations_flock;
895 extern struct file_operations ll_file_operations_noflock;
896 extern struct inode_operations ll_file_inode_operations;
897 extern int ll_have_md_lock(struct inode *inode, __u64 *bits,
898 enum ldlm_mode l_req_mode);
899 extern enum ldlm_mode ll_take_md_lock(struct inode *inode, __u64 bits,
900 struct lustre_handle *lockh, __u64 flags,
901 enum ldlm_mode mode);
903 int ll_file_open(struct inode *inode, struct file *file);
904 int ll_file_release(struct inode *inode, struct file *file);
905 int ll_release_openhandle(struct dentry *, struct lookup_intent *);
906 int ll_md_real_close(struct inode *inode, fmode_t fmode);
907 extern void ll_rw_stats_tally(struct ll_sb_info *sbi, pid_t pid,
908 struct ll_file_data *file, loff_t pos,
909 size_t count, int rw);
910 #ifdef HAVE_INODEOPS_ENHANCED_GETATTR
911 int ll_getattr(const struct path *path, struct kstat *stat,
912 u32 request_mask, unsigned int flags);
914 int ll_getattr(struct vfsmount *mnt, struct dentry *de, struct kstat *stat);
916 int ll_getattr_dentry(struct dentry *de, struct kstat *stat);
917 struct posix_acl *ll_get_acl(struct inode *inode, int type);
918 #ifdef HAVE_IOP_SET_ACL
919 #ifdef CONFIG_LUSTRE_FS_POSIX_ACL
920 int ll_set_acl(struct inode *inode, struct posix_acl *acl, int type);
921 #else /* !CONFIG_LUSTRE_FS_POSIX_ACL */
922 #define ll_set_acl NULL
923 #endif /* CONFIG_LUSTRE_FS_POSIX_ACL */
926 int ll_migrate(struct inode *parent, struct file *file,
927 struct lmv_user_md *lum, const char *name);
928 int ll_get_fid_by_name(struct inode *parent, const char *name,
929 int namelen, struct lu_fid *fid, struct inode **inode);
930 int ll_inode_permission(struct inode *inode, int mask);
931 int ll_ioctl_check_project(struct inode *inode, struct fsxattr *fa);
932 int ll_ioctl_fsgetxattr(struct inode *inode, unsigned int cmd,
934 int ll_ioctl_fssetxattr(struct inode *inode, unsigned int cmd,
937 int ll_lov_setstripe_ea_info(struct inode *inode, struct dentry *dentry,
938 __u64 flags, struct lov_user_md *lum,
940 int ll_lov_getstripe_ea_info(struct inode *inode, const char *filename,
941 struct lov_mds_md **lmm, int *lmm_size,
942 struct ptlrpc_request **request);
943 int ll_dir_setstripe(struct inode *inode, struct lov_user_md *lump,
945 int ll_dir_getstripe(struct inode *inode, void **lmmp,
946 int *lmm_size, struct ptlrpc_request **request,
948 int ll_fsync(struct file *file, loff_t start, loff_t end, int data);
949 int ll_merge_attr(const struct lu_env *env, struct inode *inode);
950 int ll_fid2path(struct inode *inode, void __user *arg);
951 int ll_data_version(struct inode *inode, __u64 *data_version, int flags);
952 int ll_hsm_release(struct inode *inode);
953 int ll_hsm_state_set(struct inode *inode, struct hsm_state_set *hss);
954 void ll_io_set_mirror(struct cl_io *io, const struct file *file);
958 int ll_d_init(struct dentry *de);
959 extern const struct dentry_operations ll_d_ops;
960 void ll_intent_drop_lock(struct lookup_intent *);
961 void ll_intent_release(struct lookup_intent *);
962 void ll_invalidate_aliases(struct inode *);
963 void ll_lookup_finish_locks(struct lookup_intent *it, struct dentry *dentry);
964 int ll_revalidate_it_finish(struct ptlrpc_request *request,
965 struct lookup_intent *it, struct dentry *de);
967 /* llite/llite_lib.c */
968 extern struct super_operations lustre_super_operations;
970 void ll_lli_init(struct ll_inode_info *lli);
971 int ll_fill_super(struct super_block *sb, struct vfsmount *mnt);
972 void ll_put_super(struct super_block *sb);
973 void ll_kill_super(struct super_block *sb);
974 struct inode *ll_inode_from_resource_lock(struct ldlm_lock *lock);
975 void ll_dir_clear_lsm_md(struct inode *inode);
976 void ll_clear_inode(struct inode *inode);
977 int ll_setattr_raw(struct dentry *dentry, struct iattr *attr,
978 enum op_xvalid xvalid, bool hsm_import);
979 int ll_setattr(struct dentry *de, struct iattr *attr);
980 int ll_statfs(struct dentry *de, struct kstatfs *sfs);
981 int ll_statfs_internal(struct ll_sb_info *sbi, struct obd_statfs *osfs,
983 int ll_update_inode(struct inode *inode, struct lustre_md *md);
984 void ll_update_inode_flags(struct inode *inode, int ext_flags);
985 int ll_read_inode2(struct inode *inode, void *opaque);
986 void ll_delete_inode(struct inode *inode);
987 int ll_iocontrol(struct inode *inode, struct file *file,
988 unsigned int cmd, unsigned long arg);
989 int ll_flush_ctx(struct inode *inode);
990 void ll_umount_begin(struct super_block *sb);
991 int ll_remount_fs(struct super_block *sb, int *flags, char *data);
992 #ifdef HAVE_SUPEROPS_USE_DENTRY
993 int ll_show_options(struct seq_file *seq, struct dentry *dentry);
995 int ll_show_options(struct seq_file *seq, struct vfsmount *vfs);
997 void ll_dirty_page_discard_warn(struct page *page, int ioret);
998 int ll_prep_inode(struct inode **inode, struct ptlrpc_request *req,
999 struct super_block *, struct lookup_intent *);
1000 int ll_obd_statfs(struct inode *inode, void __user *arg);
1001 int ll_get_max_mdsize(struct ll_sb_info *sbi, int *max_mdsize);
1002 int ll_get_default_mdsize(struct ll_sb_info *sbi, int *default_mdsize);
1003 int ll_set_default_mdsize(struct ll_sb_info *sbi, int default_mdsize);
1005 void ll_unlock_md_op_lsm(struct md_op_data *op_data);
1006 struct md_op_data *ll_prep_md_op_data(struct md_op_data *op_data,
1007 struct inode *i1, struct inode *i2,
1008 const char *name, size_t namelen,
1009 __u32 mode, enum md_op_code opc,
1011 void ll_finish_md_op_data(struct md_op_data *op_data);
1012 int ll_get_obd_name(struct inode *inode, unsigned int cmd, unsigned long arg);
1013 void ll_compute_rootsquash_state(struct ll_sb_info *sbi);
1014 ssize_t ll_copy_user_md(const struct lov_user_md __user *md,
1015 struct lov_user_md **kbuf);
1016 void ll_open_cleanup(struct super_block *sb, struct ptlrpc_request *open_req);
1018 void ll_dom_finish_open(struct inode *inode, struct ptlrpc_request *req,
1019 struct lookup_intent *it);
1021 /* Compute expected user md size when passing in a md from user space */
1022 static inline ssize_t ll_lov_user_md_size(const struct lov_user_md *lum)
1024 switch (lum->lmm_magic) {
1025 case LOV_USER_MAGIC_V1:
1026 return sizeof(struct lov_user_md_v1);
1027 case LOV_USER_MAGIC_V3:
1028 return sizeof(struct lov_user_md_v3);
1029 case LOV_USER_MAGIC_SPECIFIC:
1030 if (lum->lmm_stripe_count > LOV_MAX_STRIPE_COUNT)
1033 return lov_user_md_size(lum->lmm_stripe_count,
1034 LOV_USER_MAGIC_SPECIFIC);
1035 case LOV_USER_MAGIC_COMP_V1:
1036 return ((struct lov_comp_md_v1 *)lum)->lcm_size;
1037 case LOV_USER_MAGIC_FOREIGN:
1038 return foreign_size(lum);
1044 /* llite/llite_nfs.c */
1045 extern struct export_operations lustre_export_operations;
1046 __u32 get_uuid2int(const char *name, int len);
1047 struct inode *search_inode_for_lustre(struct super_block *sb,
1048 const struct lu_fid *fid);
1049 int ll_dir_get_parent_fid(struct inode *dir, struct lu_fid *parent_fid);
1051 /* llite/symlink.c */
1052 extern struct inode_operations ll_fast_symlink_inode_operations;
1055 * IO arguments for various VFS I/O interfaces.
1057 struct vvp_io_args {
1058 /** normal/sendfile/splice */
1059 enum vvp_io_subtype via_io_subtype;
1063 struct kiocb *via_iocb;
1064 struct iov_iter *via_iter;
1067 struct pipe_inode_info *via_pipe;
1068 unsigned int via_flags;
1078 struct ll_cl_context {
1079 struct list_head lcc_list;
1081 const struct lu_env *lcc_env;
1082 struct cl_io *lcc_io;
1083 struct cl_page *lcc_page;
1084 enum lcc_type lcc_type;
1087 struct ll_thread_info {
1088 struct vvp_io_args lti_args;
1089 struct ra_io_arg lti_ria;
1090 struct ll_cl_context lti_io_ctx;
1093 extern struct lu_context_key ll_thread_key;
1095 static inline struct ll_thread_info *ll_env_info(const struct lu_env *env)
1097 struct ll_thread_info *lti;
1099 lti = lu_context_key_get(&env->le_ctx, &ll_thread_key);
1100 LASSERT(lti != NULL);
1105 static inline struct vvp_io_args *ll_env_args(const struct lu_env *env,
1106 enum vvp_io_subtype type)
1108 struct vvp_io_args *via = &ll_env_info(env)->lti_args;
1110 via->via_io_subtype = type;
1115 void ll_io_init(struct cl_io *io, struct file *file, enum cl_io_type iot,
1116 struct vvp_io_args *args);
1118 /* llite/llite_mmap.c */
1120 int ll_teardown_mmaps(struct address_space *mapping, __u64 first, __u64 last);
1121 int ll_file_mmap(struct file * file, struct vm_area_struct * vma);
1122 void policy_from_vma(union ldlm_policy_data *policy, struct vm_area_struct *vma,
1123 unsigned long addr, size_t count);
1124 struct vm_area_struct *our_vma(struct mm_struct *mm, unsigned long addr,
1127 #define ll_s2sbi(sb) (s2lsi(sb)->lsi_llsbi)
1129 /* don't need an addref as the sb_info should be holding one */
1130 static inline struct obd_export *ll_s2dtexp(struct super_block *sb)
1132 return ll_s2sbi(sb)->ll_dt_exp;
1135 /* don't need an addref as the sb_info should be holding one */
1136 static inline struct obd_export *ll_s2mdexp(struct super_block *sb)
1138 return ll_s2sbi(sb)->ll_md_exp;
1141 static inline struct client_obd *sbi2mdc(struct ll_sb_info *sbi)
1143 struct obd_device *obd = sbi->ll_md_exp->exp_obd;
1149 // FIXME: replace the name of this with LL_SB to conform to kernel stuff
1150 static inline struct ll_sb_info *ll_i2sbi(struct inode *inode)
1152 return ll_s2sbi(inode->i_sb);
1155 static inline struct obd_export *ll_i2dtexp(struct inode *inode)
1157 return ll_s2dtexp(inode->i_sb);
1160 static inline struct obd_export *ll_i2mdexp(struct inode *inode)
1162 return ll_s2mdexp(inode->i_sb);
1165 static inline struct lu_fid *ll_inode2fid(struct inode *inode)
1169 LASSERT(inode != NULL);
1170 fid = &ll_i2info(inode)->lli_fid;
1175 static inline bool ll_dir_striped(struct inode *inode)
1178 return S_ISDIR(inode->i_mode) &&
1179 lmv_dir_striped(ll_i2info(inode)->lli_lsm_md);
1182 static inline loff_t ll_file_maxbytes(struct inode *inode)
1184 struct cl_object *obj = ll_i2info(inode)->lli_clob;
1187 return MAX_LFS_FILESIZE;
1189 return min_t(loff_t, cl_object_maxbytes(obj), MAX_LFS_FILESIZE);
1193 extern const struct xattr_handler *ll_xattr_handlers[];
1195 #define XATTR_USER_T 1
1196 #define XATTR_TRUSTED_T 2
1197 #define XATTR_SECURITY_T 3
1198 #define XATTR_ACL_ACCESS_T 4
1199 #define XATTR_ACL_DEFAULT_T 5
1200 #define XATTR_LUSTRE_T 6
1201 #define XATTR_OTHER_T 7
1203 ssize_t ll_listxattr(struct dentry *dentry, char *buffer, size_t size);
1204 int ll_xattr_list(struct inode *inode, const char *name, int type,
1205 void *buffer, size_t size, u64 valid);
1206 const struct xattr_handler *get_xattr_type(const char *name);
1209 * Common IO arguments for various VFS I/O interfaces.
1211 int cl_sb_init(struct super_block *sb);
1212 int cl_sb_fini(struct super_block *sb);
1214 enum ras_update_flags {
1218 void ll_ra_count_put(struct ll_sb_info *sbi, unsigned long len);
1219 void ll_ra_stats_inc(struct inode *inode, enum ra_stat which);
1223 #define LL_SA_RPC_MIN 2
1224 #define LL_SA_RPC_DEF 32
1225 #define LL_SA_RPC_MAX 512
1227 /* XXX: If want to support more concurrent statahead instances,
1228 * please consider to decentralize the RPC lists attached
1229 * on related import, such as imp_{sending,delayed}_list.
1231 #define LL_SA_RUNNING_MAX 256
1232 #define LL_SA_RUNNING_DEF 16
1234 #define LL_SA_CACHE_BIT 5
1235 #define LL_SA_CACHE_SIZE (1 << LL_SA_CACHE_BIT)
1236 #define LL_SA_CACHE_MASK (LL_SA_CACHE_SIZE - 1)
1238 /* per inode struct, for dir only */
1239 struct ll_statahead_info {
1240 struct dentry *sai_dentry;
1241 atomic_t sai_refcount; /* when access this struct, hold
1243 unsigned int sai_max; /* max ahead of lookup */
1244 __u64 sai_sent; /* stat requests sent count */
1245 __u64 sai_replied; /* stat requests which received
1247 __u64 sai_index; /* index of statahead entry */
1248 __u64 sai_index_wait; /* index of entry which is the
1249 * caller is waiting for */
1250 __u64 sai_hit; /* hit count */
1251 __u64 sai_miss; /* miss count:
1252 * for "ls -al" case, includes
1253 * hidden dentry miss;
1254 * for "ls -l" case, it does not
1255 * include hidden dentry miss.
1256 * "sai_miss_hidden" is used for
1259 unsigned int sai_consecutive_miss; /* consecutive miss */
1260 unsigned int sai_miss_hidden;/* "ls -al", but first dentry
1261 * is not a hidden one */
1262 unsigned int sai_skip_hidden;/* skipped hidden dentry count
1264 unsigned int sai_ls_all:1, /* "ls -al", do stat-ahead for
1266 sai_agl_valid:1,/* AGL is valid for the dir */
1267 sai_in_readpage:1;/* statahead is in readdir()*/
1268 wait_queue_head_t sai_waitq; /* stat-ahead wait queue */
1269 struct ptlrpc_thread sai_thread; /* stat-ahead thread */
1270 struct ptlrpc_thread sai_agl_thread; /* AGL thread */
1271 struct list_head sai_interim_entries; /* entries which got async
1272 * stat reply, but not
1274 struct list_head sai_entries; /* completed entries */
1275 struct list_head sai_agls; /* AGLs to be sent */
1276 struct list_head sai_cache[LL_SA_CACHE_SIZE];
1277 spinlock_t sai_cache_lock[LL_SA_CACHE_SIZE];
1278 atomic_t sai_cache_count; /* entry count in cache */
1281 int ll_statahead(struct inode *dir, struct dentry **dentry, bool unplug);
1282 void ll_authorize_statahead(struct inode *dir, void *key);
1283 void ll_deauthorize_statahead(struct inode *dir, void *key);
1286 blkcnt_t dirty_cnt(struct inode *inode);
1288 int cl_glimpse_size0(struct inode *inode, int agl);
1289 int cl_glimpse_lock(const struct lu_env *env, struct cl_io *io,
1290 struct inode *inode, struct cl_object *clob, int agl);
1292 static inline int cl_glimpse_size(struct inode *inode)
1294 return cl_glimpse_size0(inode, 0);
1297 /* AGL is 'asychronous glimpse lock', which is a speculative lock taken as
1298 * part of statahead */
1299 static inline int cl_agl(struct inode *inode)
1301 return cl_glimpse_size0(inode, 1);
1304 int ll_file_lock_ahead(struct file *file, struct llapi_lu_ladvise *ladvise);
1306 int cl_io_get(struct inode *inode, struct lu_env **envout,
1307 struct cl_io **ioout, __u16 *refcheck);
1309 static inline int ll_glimpse_size(struct inode *inode)
1311 struct ll_inode_info *lli = ll_i2info(inode);
1314 down_read(&lli->lli_glimpse_sem);
1315 rc = cl_glimpse_size(inode);
1316 lli->lli_glimpse_time = ktime_get();
1317 up_read(&lli->lli_glimpse_sem);
1321 /* dentry may statahead when statahead is enabled and current process has opened
1322 * parent directory, and this dentry hasn't accessed statahead cache before */
1324 dentry_may_statahead(struct inode *dir, struct dentry *dentry)
1326 struct ll_inode_info *lli;
1327 struct ll_dentry_data *ldd;
1329 if (ll_i2sbi(dir)->ll_sa_max == 0)
1332 lli = ll_i2info(dir);
1334 /* statahead is not allowed for this dir, there may be three causes:
1335 * 1. dir is not opened.
1336 * 2. statahead hit ratio is too low.
1337 * 3. previous stat started statahead thread failed. */
1338 if (!lli->lli_sa_enabled)
1341 /* not the same process, don't statahead */
1342 if (lli->lli_opendir_pid != current_pid())
1346 * When stating a dentry, kernel may trigger 'revalidate' or 'lookup'
1347 * multiple times, eg. for 'getattr', 'getxattr' and etc.
1348 * For patchless client, lookup intent is not accurate, which may
1349 * misguide statahead. For example:
1350 * The 'revalidate' call for 'getattr' and 'getxattr' of a dentry will
1351 * have the same intent -- IT_GETATTR, while one dentry should access
1352 * statahead cache once, otherwise statahead windows is messed up.
1353 * The solution is as following:
1354 * Assign 'lld_sa_generation' with 'lli_sa_generation' when a dentry
1355 * IT_GETATTR for the first time, and subsequent IT_GETATTR will
1356 * bypass interacting with statahead cache by checking
1357 * 'lld_sa_generation == lli->lli_sa_generation'.
1359 ldd = ll_d2d(dentry);
1360 if (ldd != NULL && ldd->lld_sa_generation == lli->lli_sa_generation)
1366 int cl_sync_file_range(struct inode *inode, loff_t start, loff_t end,
1367 enum cl_fsync_mode mode, int ignore_layout);
1369 static inline int ll_file_nolock(const struct file *file)
1371 struct ll_file_data *fd = LUSTRE_FPRIVATE(file);
1372 struct inode *inode = file_inode((struct file *)file);
1374 LASSERT(fd != NULL);
1375 return ((fd->fd_flags & LL_FILE_IGNORE_LOCK) ||
1376 (ll_i2sbi(inode)->ll_flags & LL_SBI_NOLCK));
1379 static inline void ll_set_lock_data(struct obd_export *exp, struct inode *inode,
1380 struct lookup_intent *it, __u64 *bits)
1382 if (!it->it_lock_set) {
1383 struct lustre_handle handle;
1385 /* If this inode is a remote object, it will get two
1386 * separate locks in different namespaces, Master MDT,
1387 * where the name entry is, will grant LOOKUP lock,
1388 * remote MDT, where the object is, will grant
1389 * UPDATE|PERM lock. The inode will be attched to both
1390 * LOOKUP and PERM locks, so revoking either locks will
1391 * case the dcache being cleared */
1392 if (it->it_remote_lock_mode) {
1393 handle.cookie = it->it_remote_lock_handle;
1394 CDEBUG(D_DLMTRACE, "setting l_data to inode "DFID
1395 "(%p) for remote lock %#llx\n",
1396 PFID(ll_inode2fid(inode)), inode,
1398 md_set_lock_data(exp, &handle, inode, NULL);
1401 handle.cookie = it->it_lock_handle;
1403 CDEBUG(D_DLMTRACE, "setting l_data to inode "DFID"(%p)"
1404 " for lock %#llx\n",
1405 PFID(ll_inode2fid(inode)), inode, handle.cookie);
1407 md_set_lock_data(exp, &handle, inode, &it->it_lock_bits);
1408 it->it_lock_set = 1;
1412 *bits = it->it_lock_bits;
1415 static inline int d_lustre_invalid(const struct dentry *dentry)
1417 struct ll_dentry_data *lld = ll_d2d(dentry);
1419 return (lld == NULL) || lld->lld_invalid;
1422 static inline void __d_lustre_invalidate(struct dentry *dentry)
1424 struct ll_dentry_data *lld = ll_d2d(dentry);
1427 lld->lld_invalid = 1;
1431 * Mark dentry INVALID, if dentry refcount is zero (this is normally case for
1432 * ll_md_blocking_ast), unhash this dentry, and let dcache to reclaim it later;
1433 * else dput() of the last refcount will unhash this dentry and kill it.
1435 static inline void d_lustre_invalidate(struct dentry *dentry, int nested)
1437 CDEBUG(D_DENTRY, "invalidate dentry %.*s (%p) parent %p inode %p "
1438 "refc %d\n", dentry->d_name.len, dentry->d_name.name, dentry,
1439 dentry->d_parent, dentry->d_inode, ll_d_count(dentry));
1441 spin_lock_nested(&dentry->d_lock,
1442 nested ? DENTRY_D_LOCK_NESTED : DENTRY_D_LOCK_NORMAL);
1443 __d_lustre_invalidate(dentry);
1445 * We should be careful about dentries created by d_obtain_alias().
1446 * These dentries are not put in the dentry tree, instead they are
1447 * linked to sb->s_anon through dentry->d_hash.
1448 * shrink_dcache_for_umount() shrinks the tree and sb->s_anon list.
1449 * If we unhashed such a dentry, unmount would not be able to find
1450 * it and busy inodes would be reported.
1452 if (ll_d_count(dentry) == 0 && !(dentry->d_flags & DCACHE_DISCONNECTED))
1454 spin_unlock(&dentry->d_lock);
1457 static inline void d_lustre_revalidate(struct dentry *dentry)
1459 spin_lock(&dentry->d_lock);
1460 LASSERT(ll_d2d(dentry) != NULL);
1461 ll_d2d(dentry)->lld_invalid = 0;
1462 spin_unlock(&dentry->d_lock);
1465 static inline dev_t ll_compat_encode_dev(dev_t dev)
1467 /* The compat_sys_*stat*() syscalls will fail unless the
1468 * device majors and minors are both less than 256. Note that
1469 * the value returned here will be passed through
1470 * old_encode_dev() in cp_compat_stat(). And so we are not
1471 * trying to return a valid compat (u16) device number, just
1472 * one that will pass the old_valid_dev() check. */
1474 return MKDEV(MAJOR(dev) & 0xff, MINOR(dev) & 0xff);
1477 int ll_layout_conf(struct inode *inode, const struct cl_object_conf *conf);
1478 int ll_layout_refresh(struct inode *inode, __u32 *gen);
1479 int ll_layout_restore(struct inode *inode, loff_t start, __u64 length);
1480 int ll_layout_write_intent(struct inode *inode, enum layout_intent_opc opc,
1481 struct lu_extent *ext);
1483 int ll_xattr_init(void);
1484 void ll_xattr_fini(void);
1486 int ll_page_sync_io(const struct lu_env *env, struct cl_io *io,
1487 struct cl_page *page, enum cl_req_type crt);
1489 int ll_getparent(struct file *file, struct getparent __user *arg);
1492 int cl_setattr_ost(struct cl_object *obj, const struct iattr *attr,
1493 enum op_xvalid xvalid, unsigned int attr_flags);
1495 extern struct lu_env *cl_inode_fini_env;
1496 extern __u16 cl_inode_fini_refcheck;
1498 int cl_file_inode_init(struct inode *inode, struct lustre_md *md);
1499 void cl_inode_fini(struct inode *inode);
1501 u64 cl_fid_build_ino(const struct lu_fid *fid, int api32);
1502 u32 cl_fid_build_gen(const struct lu_fid *fid);
1504 #endif /* LLITE_INTERNAL_H */