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 /* Try to make the d::member and f::member are aligned. Before using
138 * these members, make clear whether it is directory or not. */
142 /* metadata statahead */
143 /* since parent-child threads can share the same @file
144 * struct, "opendir_key" is the token when dir close for
145 * case of parent exit before child -- it is me should
146 * cleanup the dir readahead. */
147 void *lli_opendir_key;
148 struct ll_statahead_info *lli_sai;
149 /* protect statahead stuff. */
150 spinlock_t lli_sa_lock;
151 /* "opendir_pid" is the token when lookup/revalid
152 * -- I am the owner of dir statahead. */
153 pid_t lli_opendir_pid;
154 /* stat will try to access statahead entries or start
155 * statahead if this flag is set, and this flag will be
156 * set upon dir open, and cleared when dir is closed,
157 * statahead hit ratio is too low, or start statahead
159 unsigned int lli_sa_enabled:1;
160 /* generation for statahead */
161 unsigned int lli_sa_generation;
162 /* rw lock protects lli_lsm_md */
163 struct rw_semaphore lli_lsm_sem;
164 /* directory stripe information */
165 struct lmv_stripe_md *lli_lsm_md;
166 /* directory default LMV */
167 struct lmv_stripe_md *lli_default_lsm_md;
170 /* for non-directory */
172 struct mutex lli_size_mutex;
173 char *lli_symlink_name;
175 * struct rw_semaphore {
176 * signed long count; // align d.d_def_acl
177 * spinlock_t wait_lock; // align d.d_sa_lock
178 * struct list_head wait_list;
181 struct rw_semaphore lli_trunc_sem;
182 struct range_lock_tree lli_write_tree;
184 struct rw_semaphore lli_glimpse_sem;
185 ktime_t lli_glimpse_time;
186 struct list_head lli_agl_list;
189 /* for writepage() only to communicate to fsync */
192 /* protect the file heat fields */
193 spinlock_t lli_heat_lock;
194 __u32 lli_heat_flags;
195 struct obd_heat_instance lli_heat_instances[OBD_HEAT_COUNT];
198 * Whenever a process try to read/write the file, the
199 * jobid of the process will be saved here, and it'll
200 * be packed into the write PRC when flush later.
202 * So the read/write statistics for jobid will not be
203 * accurate if the file is shared by different jobs.
205 char lli_jobid[LUSTRE_JOBID_SIZE];
207 struct mutex lli_pcc_lock;
208 enum lu_pcc_state_flags lli_pcc_state;
210 * @lli_pcc_generation saves the gobal PCC generation
211 * when the file was successfully attached into PCC.
212 * The flags of the PCC dataset are saved in
214 * The gobal PCC generation will be increased when add
215 * or delete a PCC backend, or change the configuration
216 * parameters for PCC.
217 * If @lli_pcc_generation is same as the gobal PCC
218 * generation, we can use the saved flags of the PCC
219 * dataset to determine whether need to try auto attach
222 __u64 lli_pcc_generation;
223 enum pcc_dataset_flags lli_pcc_dsflags;
224 struct pcc_inode *lli_pcc_inode;
226 struct mutex lli_group_mutex;
227 __u64 lli_group_users;
228 unsigned long lli_group_gid;
232 /* XXX: For following frequent used members, although they maybe special
233 * used for non-directory object, it is some time-wasting to check
234 * whether the object is directory or not before using them. On the
235 * other hand, currently, sizeof(f) > sizeof(d), it cannot reduce
236 * the "ll_inode_info" size even if moving those members into u.f.
237 * So keep them out side.
239 * In the future, if more members are added only for directory,
240 * some of the following members can be moved into u.f.
242 struct cl_object *lli_clob;
244 /* mutex to request for layout lock exclusively. */
245 struct mutex lli_layout_mutex;
246 /* Layout version, protected by lli_layout_lock */
247 __u32 lli_layout_gen;
248 spinlock_t lli_layout_lock;
250 __u32 lli_projid; /* project id */
252 struct rw_semaphore lli_xattrs_list_rwsem;
253 struct mutex lli_xattrs_enq_lock;
254 struct list_head lli_xattrs; /* ll_xattr_entry->xe_list */
257 static inline __u32 ll_layout_version_get(struct ll_inode_info *lli)
261 spin_lock(&lli->lli_layout_lock);
262 gen = lli->lli_layout_gen;
263 spin_unlock(&lli->lli_layout_lock);
268 static inline void ll_layout_version_set(struct ll_inode_info *lli, __u32 gen)
270 spin_lock(&lli->lli_layout_lock);
271 lli->lli_layout_gen = gen;
272 spin_unlock(&lli->lli_layout_lock);
276 /* File data is modified. */
277 LLIF_DATA_MODIFIED = 0,
278 /* File is being restored */
279 LLIF_FILE_RESTORING = 1,
280 /* Xattr cache is attached to the file */
281 LLIF_XATTR_CACHE = 2,
282 /* Project inherit */
283 LLIF_PROJECT_INHERIT = 3,
284 /* update atime from MDS even if it's older than local inode atime. */
285 LLIF_UPDATE_ATIME = 4,
289 static inline void ll_file_set_flag(struct ll_inode_info *lli,
290 enum ll_file_flags flag)
292 set_bit(flag, &lli->lli_flags);
295 static inline void ll_file_clear_flag(struct ll_inode_info *lli,
296 enum ll_file_flags flag)
298 clear_bit(flag, &lli->lli_flags);
301 static inline bool ll_file_test_flag(struct ll_inode_info *lli,
302 enum ll_file_flags flag)
304 return test_bit(flag, &lli->lli_flags);
307 static inline bool ll_file_test_and_clear_flag(struct ll_inode_info *lli,
308 enum ll_file_flags flag)
310 return test_and_clear_bit(flag, &lli->lli_flags);
313 int ll_xattr_cache_destroy(struct inode *inode);
315 int ll_xattr_cache_get(struct inode *inode,
321 static inline bool obd_connect_has_secctx(struct obd_connect_data *data)
323 #if defined(HAVE_SECURITY_DENTRY_INIT_SECURITY) && defined(CONFIG_SECURITY)
324 return data->ocd_connect_flags & OBD_CONNECT_FLAGS2 &&
325 data->ocd_connect_flags2 & OBD_CONNECT2_FILE_SECCTX;
328 #endif /* HAVE_SECURITY_DENTRY_INIT_SECURITY */
331 static inline void obd_connect_set_secctx(struct obd_connect_data *data)
333 #if defined(HAVE_SECURITY_DENTRY_INIT_SECURITY) && defined(CONFIG_SECURITY)
334 data->ocd_connect_flags2 |= OBD_CONNECT2_FILE_SECCTX;
335 #endif /* HAVE_SECURITY_DENTRY_INIT_SECURITY */
338 int ll_dentry_init_security(struct dentry *dentry, int mode, struct qstr *name,
339 const char **secctx_name, void **secctx,
341 int ll_inode_init_security(struct dentry *dentry, struct inode *inode,
344 int ll_listsecurity(struct inode *inode, char *secctx_name,
345 size_t secctx_name_size);
348 * Locking to guarantee consistency of non-atomic updates to long long i_size,
349 * consistency between file size and KMS.
351 * Implemented by ->lli_size_mutex and ->lsm_lock, nested in that order.
354 void ll_inode_size_lock(struct inode *inode);
355 void ll_inode_size_unlock(struct inode *inode);
357 static inline struct ll_inode_info *ll_i2info(struct inode *inode)
359 return container_of(inode, struct ll_inode_info, lli_vfs_inode);
362 static inline struct pcc_inode *ll_i2pcci(struct inode *inode)
364 return ll_i2info(inode)->lli_pcc_inode;
367 /* default to use at least 16M for fast read if possible */
368 #define RA_REMAIN_WINDOW_MIN MiB_TO_PAGES(16UL)
370 /* default to about 64M of readahead on a given system. */
371 #define SBI_DEFAULT_READAHEAD_MAX MiB_TO_PAGES(64UL)
373 /* default to read-ahead full files smaller than 2MB on the second read */
374 #define SBI_DEFAULT_READAHEAD_WHOLE_MAX MiB_TO_PAGES(2UL)
379 RA_STAT_DISTANT_READPAGE,
380 RA_STAT_MISS_IN_WINDOW,
381 RA_STAT_FAILED_GRAB_PAGE,
382 RA_STAT_FAILED_MATCH,
387 RA_STAT_MAX_IN_FLIGHT,
388 RA_STAT_WRONG_GRAB_PAGE,
389 RA_STAT_FAILED_REACH_END,
391 RA_STAT_FAILED_FAST_READ,
396 atomic_t ra_cur_pages;
397 unsigned long ra_max_pages;
398 unsigned long ra_max_pages_per_file;
399 unsigned long ra_max_read_ahead_whole_pages;
400 struct workqueue_struct *ll_readahead_wq;
402 * Max number of active works for readahead workqueue,
403 * default is 0 which make workqueue init number itself,
404 * unless there is a specific need for throttling the
405 * number of active work items, specifying '0' is recommended.
407 unsigned int ra_async_max_active;
408 /* Threshold to control when to trigger async readahead */
409 unsigned long ra_async_pages_per_file_threshold;
412 /* ra_io_arg will be filled in the beginning of ll_readahead with
413 * ras_lock, then the following ll_read_ahead_pages will read RA
414 * pages according to this arg, all the items in this structure are
415 * counted by page index.
418 pgoff_t ria_start; /* start offset of read-ahead*/
419 pgoff_t ria_end; /* end offset of read-ahead*/
420 unsigned long ria_reserved; /* reserved pages for read-ahead */
421 pgoff_t ria_end_min; /* minimum end to cover current read */
422 bool ria_eof; /* reach end of file */
423 /* If stride read pattern is detected, ria_stoff means where
424 * stride read is started. Note: for normal read-ahead, the
425 * value here is meaningless, and also it will not be accessed*/
426 unsigned long ria_stoff;
427 /* ria_length and ria_bytes are the length and pages length in the
428 * stride I/O mode. And they will also be used to check whether
429 * it is stride I/O read-ahead in the read-ahead pages*/
430 unsigned long ria_length;
431 unsigned long ria_bytes;
434 /* LL_HIST_MAX=32 causes an overflow */
435 #define LL_HIST_MAX 28
436 #define LL_HIST_START 12 /* buckets start at 2^12 = 4k */
437 #define LL_PROCESS_HIST_MAX 10
438 struct per_process_info {
440 struct obd_histogram pp_r_hist;
441 struct obd_histogram pp_w_hist;
444 /* pp_extents[LL_PROCESS_HIST_MAX] will hold the combined process info */
445 struct ll_rw_extents_info {
446 struct per_process_info pp_extents[LL_PROCESS_HIST_MAX + 1];
449 #define LL_OFFSET_HIST_MAX 100
450 struct ll_rw_process_info {
453 loff_t rw_range_start;
455 loff_t rw_last_file_pos;
457 size_t rw_smallest_extent;
458 size_t rw_largest_extent;
459 struct ll_file_data *rw_last_file;
462 enum stats_track_type {
463 STATS_TRACK_ALL = 0, /* track all processes */
464 STATS_TRACK_PID, /* track process with this pid */
465 STATS_TRACK_PPID, /* track processes with this ppid */
466 STATS_TRACK_GID, /* track processes with this gid */
470 /* flags for sbi->ll_flags */
471 #define LL_SBI_NOLCK 0x01 /* DLM locking disabled (directio-only) */
472 #define LL_SBI_CHECKSUM 0x02 /* checksum each page as it's written */
473 #define LL_SBI_FLOCK 0x04
474 #define LL_SBI_USER_XATTR 0x08 /* support user xattr */
475 #define LL_SBI_ACL 0x10 /* support ACL */
476 /* LL_SBI_RMT_CLIENT 0x40 remote client */
477 #define LL_SBI_MDS_CAPA 0x80 /* support mds capa, obsolete */
478 #define LL_SBI_OSS_CAPA 0x100 /* support oss capa, obsolete */
479 #define LL_SBI_LOCALFLOCK 0x200 /* Local flocks support by kernel */
480 #define LL_SBI_LRU_RESIZE 0x400 /* lru resize support */
481 #define LL_SBI_LAZYSTATFS 0x800 /* lazystatfs mount option */
482 /* LL_SBI_SOM_PREVIEW 0x1000 SOM preview mount option, obsolete */
483 #define LL_SBI_32BIT_API 0x2000 /* generate 32 bit inodes. */
484 #define LL_SBI_64BIT_HASH 0x4000 /* support 64-bits dir hash/offset */
485 #define LL_SBI_AGL_ENABLED 0x8000 /* enable agl */
486 #define LL_SBI_VERBOSE 0x10000 /* verbose mount/umount */
487 #define LL_SBI_LAYOUT_LOCK 0x20000 /* layout lock support */
488 #define LL_SBI_USER_FID2PATH 0x40000 /* allow fid2path by unprivileged users */
489 #define LL_SBI_XATTR_CACHE 0x80000 /* support for xattr cache */
490 #define LL_SBI_NOROOTSQUASH 0x100000 /* do not apply root squash */
491 #define LL_SBI_ALWAYS_PING 0x200000 /* always ping even if server
493 #define LL_SBI_FAST_READ 0x400000 /* fast read support */
494 #define LL_SBI_FILE_SECCTX 0x800000 /* set file security context at create */
495 /* LL_SBI_PIO 0x1000000 parallel IO support, introduced in
497 #define LL_SBI_TINY_WRITE 0x2000000 /* tiny write support */
498 #define LL_SBI_FILE_HEAT 0x4000000 /* file heat support */
499 #define LL_SBI_FLAGS { \
529 /* This is embedded into llite super-blocks to keep track of connect
530 * flags (capabilities) supported by all imports given mount is
532 struct lustre_client_ocd {
533 /* This is conjunction of connect_flags across all imports
534 * (LOVs) this mount is connected to. This field is updated by
535 * cl_ocd_update() under ->lco_lock. */
537 struct mutex lco_lock;
538 struct obd_export *lco_md_exp;
539 struct obd_export *lco_dt_exp;
543 /* this protects pglist and ra_info. It isn't safe to
544 * grab from interrupt contexts */
546 spinlock_t ll_pp_extent_lock; /* pp_extent entry*/
547 spinlock_t ll_process_lock; /* ll_rw_process_info */
548 struct obd_uuid ll_sb_uuid;
549 struct obd_export *ll_md_exp;
550 struct obd_export *ll_dt_exp;
551 struct obd_device *ll_md_obd;
552 struct obd_device *ll_dt_obd;
553 struct dentry *ll_debugfs_entry;
554 struct lu_fid ll_root_fid; /* root object fid */
557 unsigned int ll_xattr_cache_enabled:1,
558 ll_xattr_cache_set:1, /* already set to 0/1 */
559 ll_client_common_fill_super_succeeded:1,
562 struct lustre_client_ocd ll_lco;
564 struct lprocfs_stats *ll_stats; /* lprocfs stats counter */
566 /* Used to track "unstable" pages on a client, and maintain a
567 * LRU list of clean pages. An "unstable" page is defined as
568 * any page which is sent to a server as part of a bulk request,
569 * but is uncommitted to stable storage. */
570 struct cl_client_cache *ll_cache;
572 struct lprocfs_stats *ll_ra_stats;
574 struct ll_ra_info ll_ra_info;
575 unsigned int ll_namelen;
576 struct file_operations *ll_fop;
578 struct lu_site *ll_site;
579 struct cl_device *ll_cl;
581 struct ll_rw_extents_info ll_rw_extents_info;
582 int ll_extent_process_count;
583 struct ll_rw_process_info ll_rw_process_info[LL_PROCESS_HIST_MAX];
584 unsigned int ll_offset_process_count;
585 struct ll_rw_process_info ll_rw_offset_info[LL_OFFSET_HIST_MAX];
586 unsigned int ll_rw_offset_entry_count;
587 int ll_stats_track_id;
588 enum stats_track_type ll_stats_track_type;
591 /* metadata stat-ahead */
592 unsigned int ll_sa_running_max;/* max concurrent
593 * statahead instances */
594 unsigned int ll_sa_max; /* max statahead RPCs */
595 atomic_t ll_sa_total; /* statahead thread started
597 atomic_t ll_sa_wrong; /* statahead thread stopped for
599 atomic_t ll_sa_running; /* running statahead thread
601 atomic_t ll_agl_total; /* AGL thread started count */
603 dev_t ll_sdev_orig; /* save s_dev before assign for
606 struct root_squash_info ll_squash;
609 /* st_blksize returned by stat(2), when non-zero */
610 unsigned int ll_stat_blksize;
612 /* maximum relative age of cached statfs results */
613 unsigned int ll_statfs_max_age;
615 struct kset ll_kset; /* sysfs object */
616 struct completion ll_kobj_unregister;
619 unsigned int ll_heat_decay_weight;
620 unsigned int ll_heat_period_second;
622 /* filesystem fsname */
623 char ll_fsname[LUSTRE_MAXFSNAME + 1];
625 /* Persistent Client Cache */
626 struct pcc_super ll_pcc_super;
629 #define SBI_DEFAULT_HEAT_DECAY_WEIGHT ((80 * 256 + 50) / 100)
630 #define SBI_DEFAULT_HEAT_PERIOD_SECOND (60)
632 * per file-descriptor read-ahead data.
634 struct ll_readahead_state {
636 /* End byte that read(2) try to read. */
637 unsigned long ras_last_read_end;
639 * number of bytes read after last read-ahead window reset. As window
640 * is reset on each seek, this is effectively a number of consecutive
641 * accesses. Maybe ->ras_accessed_in_window is better name.
643 * XXX nikita: window is also reset (by ras_update()) when Lustre
644 * believes that memory pressure evicts read-ahead pages. In that
645 * case, it probably doesn't make sense to expand window to
646 * PTLRPC_MAX_BRW_PAGES on the third access.
648 unsigned long ras_consecutive_bytes;
650 * number of read requests after the last read-ahead window reset
651 * As window is reset on each seek, this is effectively the number
652 * on consecutive read request and is used to trigger read-ahead.
654 unsigned long ras_consecutive_requests;
656 * Parameters of current read-ahead window. Handled by
657 * ras_update(). On the initial access to the file or after a seek,
658 * window is reset to 0. After 3 consecutive accesses, window is
659 * expanded to PTLRPC_MAX_BRW_PAGES. Afterwards, window is enlarged by
660 * PTLRPC_MAX_BRW_PAGES chunks up to ->ra_max_pages.
662 pgoff_t ras_window_start, ras_window_len;
664 * Optimal RPC size. It decides how many pages will be sent
665 * for each read-ahead.
667 unsigned long ras_rpc_size;
669 * Where next read-ahead should start at. This lies within read-ahead
670 * window. Read-ahead window is read in pieces rather than at once
671 * because: 1. lustre limits total number of pages under read-ahead by
672 * ->ra_max_pages (see ll_ra_count_get()), 2. client cannot read pages
673 * not covered by DLM lock.
675 pgoff_t ras_next_readahead;
677 * Total number of ll_file_read requests issued, reads originating
678 * due to mmap are not counted in this total. This value is used to
679 * trigger full file read-ahead after multiple reads to a small file.
681 unsigned long ras_requests;
683 * The following 3 items are used for detecting the stride I/O
685 * In stride I/O mode,
686 * ...............|-----data-----|****gap*****|--------|******|....
687 * offset |-stride_bytes-|-stride_gap-|
688 * ras_stride_offset = offset;
689 * ras_stride_length = stride_bytes + stride_gap;
690 * ras_stride_bytes = stride_bytes;
691 * Note: all these three items are counted by bytes.
693 unsigned long ras_stride_length;
694 unsigned long ras_stride_bytes;
695 unsigned long ras_stride_offset;
697 * number of consecutive stride request count, and it is similar as
698 * ras_consecutive_requests, but used for stride I/O mode.
699 * Note: only more than 2 consecutive stride request are detected,
700 * stride read-ahead will be enable
702 unsigned long ras_consecutive_stride_requests;
703 /* index of the last page that async readahead starts */
704 pgoff_t ras_async_last_readpage;
705 /* whether we should increase readahead window */
706 bool ras_need_increase_window;
707 /* whether ra miss check should be skipped */
708 bool ras_no_miss_check;
711 struct ll_readahead_work {
712 /** File to readahead */
713 struct file *lrw_file;
715 unsigned long lrw_start;
717 unsigned long lrw_end;
719 /* async worker to handler read */
720 struct work_struct lrw_readahead_work;
723 extern struct kmem_cache *ll_file_data_slab;
724 struct lustre_handle;
725 struct ll_file_data {
726 struct ll_readahead_state fd_ras;
727 struct ll_grouplock fd_grouplock;
731 /* openhandle if lease exists for this file.
732 * Borrow lli->lli_och_mutex to protect assignment */
733 struct obd_client_handle *fd_lease_och;
734 struct obd_client_handle *fd_och;
735 struct file *fd_file;
736 /* Indicate whether need to report failure when close.
737 * true: failure is known, not report again.
738 * false: unknown failure, should report. */
739 bool fd_write_failed;
740 bool ll_lock_no_expand;
741 rwlock_t fd_lock; /* protect lcc list */
742 struct list_head fd_lccs; /* list of ll_cl_context */
743 /* Used by mirrored file to lead IOs to a specific mirror, usually
744 * for mirror resync. 0 means default. */
745 __u32 fd_designated_mirror;
746 /* The layout version when resync starts. Resync I/O should carry this
747 * layout version for verification to OST objects */
748 __u32 fd_layout_version;
749 struct pcc_file fd_pcc_file;
752 void llite_tunables_unregister(void);
753 int llite_tunables_register(void);
755 static inline struct inode *ll_info2i(struct ll_inode_info *lli)
757 return &lli->lli_vfs_inode;
760 __u32 ll_i2suppgid(struct inode *i);
761 void ll_i2gids(__u32 *suppgids, struct inode *i1,struct inode *i2);
763 static inline int ll_need_32bit_api(struct ll_sb_info *sbi)
765 #if BITS_PER_LONG == 32
767 #elif defined(CONFIG_COMPAT)
768 if (unlikely(sbi->ll_flags & LL_SBI_32BIT_API))
771 # ifdef CONFIG_X86_X32
772 /* in_compat_syscall() returns true when called from a kthread
773 * and CONFIG_X86_X32 is enabled, which is wrong. So check
774 * whether the caller comes from a syscall (ie. not a kthread)
775 * before calling in_compat_syscall(). */
776 if (current->flags & PF_KTHREAD)
780 return unlikely(in_compat_syscall());
782 return unlikely(sbi->ll_flags & LL_SBI_32BIT_API);
786 static inline bool ll_sbi_has_fast_read(struct ll_sb_info *sbi)
788 return !!(sbi->ll_flags & LL_SBI_FAST_READ);
791 static inline bool ll_sbi_has_tiny_write(struct ll_sb_info *sbi)
793 return !!(sbi->ll_flags & LL_SBI_TINY_WRITE);
796 static inline bool ll_sbi_has_file_heat(struct ll_sb_info *sbi)
798 return !!(sbi->ll_flags & LL_SBI_FILE_HEAT);
801 void ll_ras_enter(struct file *f, unsigned long pos, unsigned long count);
803 /* llite/lcommon_misc.c */
804 int cl_ocd_update(struct obd_device *host, struct obd_device *watched,
805 enum obd_notify_event ev, void *owner);
806 int cl_get_grouplock(struct cl_object *obj, unsigned long gid, int nonblock,
807 struct ll_grouplock *lg);
808 void cl_put_grouplock(struct ll_grouplock *lg);
810 /* llite/lproc_llite.c */
811 int ll_debugfs_register_super(struct super_block *sb, const char *name);
812 void ll_debugfs_unregister_super(struct super_block *sb);
813 void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, long count);
817 LPROC_LL_WRITE_BYTES,
844 LPROC_LL_GETXATTR_HITS,
846 LPROC_LL_REMOVEXATTR,
848 LPROC_LL_FILE_OPCODES
852 struct ll_dir_chain {
855 static inline void ll_dir_chain_init(struct ll_dir_chain *chain)
859 static inline void ll_dir_chain_fini(struct ll_dir_chain *chain)
863 extern const struct file_operations ll_dir_operations;
864 extern const struct inode_operations ll_dir_inode_operations;
865 #ifdef HAVE_DIR_CONTEXT
866 int ll_dir_read(struct inode *inode, __u64 *pos, struct md_op_data *op_data,
867 struct dir_context *ctx);
869 int ll_dir_read(struct inode *inode, __u64 *pos, struct md_op_data *op_data,
870 void *cookie, filldir_t filldir);
872 int ll_get_mdt_idx(struct inode *inode);
873 int ll_get_mdt_idx_by_fid(struct ll_sb_info *sbi, const struct lu_fid *fid);
874 struct page *ll_get_dir_page(struct inode *dir, struct md_op_data *op_data,
875 __u64 offset, struct ll_dir_chain *chain);
876 void ll_release_page(struct inode *inode, struct page *page, bool remove);
879 extern const struct inode_operations ll_special_inode_operations;
881 struct inode *ll_iget(struct super_block *sb, ino_t hash,
882 struct lustre_md *lic);
883 int ll_test_inode_by_fid(struct inode *inode, void *opaque);
884 int ll_md_blocking_ast(struct ldlm_lock *, struct ldlm_lock_desc *,
885 void *data, int flag);
886 struct dentry *ll_splice_alias(struct inode *inode, struct dentry *de);
887 int ll_rmdir_entry(struct inode *dir, char *name, int namelen);
888 void ll_update_times(struct ptlrpc_request *request, struct inode *inode);
891 int ll_writepage(struct page *page, struct writeback_control *wbc);
892 int ll_writepages(struct address_space *, struct writeback_control *wbc);
893 int ll_readpage(struct file *file, struct page *page);
894 int ll_io_read_page(const struct lu_env *env, struct cl_io *io,
895 struct cl_page *page, struct file *file);
896 void ll_readahead_init(struct inode *inode, struct ll_readahead_state *ras);
897 int vvp_io_write_commit(const struct lu_env *env, struct cl_io *io);
900 void ll_cl_add(struct file *file, const struct lu_env *env, struct cl_io *io,
902 void ll_cl_remove(struct file *file, const struct lu_env *env);
903 struct ll_cl_context *ll_cl_find(struct file *file);
905 extern const struct address_space_operations ll_aops;
908 extern struct file_operations ll_file_operations;
909 extern struct file_operations ll_file_operations_flock;
910 extern struct file_operations ll_file_operations_noflock;
911 extern struct inode_operations ll_file_inode_operations;
912 extern int ll_have_md_lock(struct inode *inode, __u64 *bits,
913 enum ldlm_mode l_req_mode);
914 extern enum ldlm_mode ll_take_md_lock(struct inode *inode, __u64 bits,
915 struct lustre_handle *lockh, __u64 flags,
916 enum ldlm_mode mode);
918 int ll_file_open(struct inode *inode, struct file *file);
919 int ll_file_release(struct inode *inode, struct file *file);
920 int ll_release_openhandle(struct dentry *, struct lookup_intent *);
921 int ll_md_real_close(struct inode *inode, fmode_t fmode);
922 extern void ll_rw_stats_tally(struct ll_sb_info *sbi, pid_t pid,
923 struct ll_file_data *file, loff_t pos,
924 size_t count, int rw);
925 #ifdef HAVE_INODEOPS_ENHANCED_GETATTR
926 int ll_getattr(const struct path *path, struct kstat *stat,
927 u32 request_mask, unsigned int flags);
929 int ll_getattr(struct vfsmount *mnt, struct dentry *de, struct kstat *stat);
931 int ll_getattr_dentry(struct dentry *de, struct kstat *stat);
932 struct posix_acl *ll_get_acl(struct inode *inode, int type);
933 #ifdef HAVE_IOP_SET_ACL
934 #ifdef CONFIG_LUSTRE_FS_POSIX_ACL
935 int ll_set_acl(struct inode *inode, struct posix_acl *acl, int type);
936 #else /* !CONFIG_LUSTRE_FS_POSIX_ACL */
937 #define ll_set_acl NULL
938 #endif /* CONFIG_LUSTRE_FS_POSIX_ACL */
941 int ll_migrate(struct inode *parent, struct file *file,
942 struct lmv_user_md *lum, const char *name);
943 int ll_get_fid_by_name(struct inode *parent, const char *name,
944 int namelen, struct lu_fid *fid, struct inode **inode);
945 int ll_inode_permission(struct inode *inode, int mask);
946 int ll_ioctl_check_project(struct inode *inode, struct fsxattr *fa);
947 int ll_ioctl_fsgetxattr(struct inode *inode, unsigned int cmd,
949 int ll_ioctl_fssetxattr(struct inode *inode, unsigned int cmd,
952 int ll_lov_setstripe_ea_info(struct inode *inode, struct dentry *dentry,
953 __u64 flags, struct lov_user_md *lum,
955 int ll_lov_getstripe_ea_info(struct inode *inode, const char *filename,
956 struct lov_mds_md **lmm, int *lmm_size,
957 struct ptlrpc_request **request);
958 int ll_dir_setstripe(struct inode *inode, struct lov_user_md *lump,
960 int ll_dir_getstripe(struct inode *inode, void **lmmp,
961 int *lmm_size, struct ptlrpc_request **request,
963 int ll_fsync(struct file *file, loff_t start, loff_t end, int data);
964 int ll_merge_attr(const struct lu_env *env, struct inode *inode);
965 int ll_fid2path(struct inode *inode, void __user *arg);
966 int ll_data_version(struct inode *inode, __u64 *data_version, int flags);
967 int ll_hsm_release(struct inode *inode);
968 int ll_hsm_state_set(struct inode *inode, struct hsm_state_set *hss);
969 void ll_io_set_mirror(struct cl_io *io, const struct file *file);
973 int ll_d_init(struct dentry *de);
974 extern const struct dentry_operations ll_d_ops;
975 void ll_intent_drop_lock(struct lookup_intent *);
976 void ll_intent_release(struct lookup_intent *);
977 void ll_invalidate_aliases(struct inode *);
978 void ll_lookup_finish_locks(struct lookup_intent *it, struct dentry *dentry);
979 int ll_revalidate_it_finish(struct ptlrpc_request *request,
980 struct lookup_intent *it, struct dentry *de);
982 /* llite/llite_lib.c */
983 extern struct super_operations lustre_super_operations;
985 void ll_lli_init(struct ll_inode_info *lli);
986 int ll_fill_super(struct super_block *sb, struct vfsmount *mnt);
987 void ll_put_super(struct super_block *sb);
988 void ll_kill_super(struct super_block *sb);
989 struct inode *ll_inode_from_resource_lock(struct ldlm_lock *lock);
990 void ll_dir_clear_lsm_md(struct inode *inode);
991 void ll_clear_inode(struct inode *inode);
992 int ll_setattr_raw(struct dentry *dentry, struct iattr *attr,
993 enum op_xvalid xvalid, bool hsm_import);
994 int ll_setattr(struct dentry *de, struct iattr *attr);
995 int ll_statfs(struct dentry *de, struct kstatfs *sfs);
996 int ll_statfs_internal(struct ll_sb_info *sbi, struct obd_statfs *osfs,
998 int ll_update_inode(struct inode *inode, struct lustre_md *md);
999 void ll_update_inode_flags(struct inode *inode, int ext_flags);
1000 int ll_read_inode2(struct inode *inode, void *opaque);
1001 void ll_delete_inode(struct inode *inode);
1002 int ll_iocontrol(struct inode *inode, struct file *file,
1003 unsigned int cmd, unsigned long arg);
1004 int ll_flush_ctx(struct inode *inode);
1005 void ll_umount_begin(struct super_block *sb);
1006 int ll_remount_fs(struct super_block *sb, int *flags, char *data);
1007 #ifdef HAVE_SUPEROPS_USE_DENTRY
1008 int ll_show_options(struct seq_file *seq, struct dentry *dentry);
1010 int ll_show_options(struct seq_file *seq, struct vfsmount *vfs);
1012 void ll_dirty_page_discard_warn(struct page *page, int ioret);
1013 int ll_prep_inode(struct inode **inode, struct ptlrpc_request *req,
1014 struct super_block *, struct lookup_intent *);
1015 int ll_obd_statfs(struct inode *inode, void __user *arg);
1016 int ll_get_max_mdsize(struct ll_sb_info *sbi, int *max_mdsize);
1017 int ll_get_default_mdsize(struct ll_sb_info *sbi, int *default_mdsize);
1018 int ll_set_default_mdsize(struct ll_sb_info *sbi, int default_mdsize);
1020 void ll_unlock_md_op_lsm(struct md_op_data *op_data);
1021 struct md_op_data *ll_prep_md_op_data(struct md_op_data *op_data,
1022 struct inode *i1, struct inode *i2,
1023 const char *name, size_t namelen,
1024 __u32 mode, enum md_op_code opc,
1026 void ll_finish_md_op_data(struct md_op_data *op_data);
1027 int ll_get_obd_name(struct inode *inode, unsigned int cmd, unsigned long arg);
1028 void ll_compute_rootsquash_state(struct ll_sb_info *sbi);
1029 ssize_t ll_copy_user_md(const struct lov_user_md __user *md,
1030 struct lov_user_md **kbuf);
1031 void ll_open_cleanup(struct super_block *sb, struct ptlrpc_request *open_req);
1033 void ll_dom_finish_open(struct inode *inode, struct ptlrpc_request *req,
1034 struct lookup_intent *it);
1036 /* Compute expected user md size when passing in a md from user space */
1037 static inline ssize_t ll_lov_user_md_size(const struct lov_user_md *lum)
1039 switch (lum->lmm_magic) {
1040 case LOV_USER_MAGIC_V1:
1041 return sizeof(struct lov_user_md_v1);
1042 case LOV_USER_MAGIC_V3:
1043 return sizeof(struct lov_user_md_v3);
1044 case LOV_USER_MAGIC_SPECIFIC:
1045 if (lum->lmm_stripe_count > LOV_MAX_STRIPE_COUNT)
1048 return lov_user_md_size(lum->lmm_stripe_count,
1049 LOV_USER_MAGIC_SPECIFIC);
1050 case LOV_USER_MAGIC_COMP_V1:
1051 return ((struct lov_comp_md_v1 *)lum)->lcm_size;
1052 case LOV_USER_MAGIC_FOREIGN:
1053 return foreign_size(lum);
1059 /* llite/llite_nfs.c */
1060 extern struct export_operations lustre_export_operations;
1061 __u32 get_uuid2int(const char *name, int len);
1062 struct inode *search_inode_for_lustre(struct super_block *sb,
1063 const struct lu_fid *fid);
1064 int ll_dir_get_parent_fid(struct inode *dir, struct lu_fid *parent_fid);
1066 /* llite/symlink.c */
1067 extern struct inode_operations ll_fast_symlink_inode_operations;
1070 * IO arguments for various VFS I/O interfaces.
1072 struct vvp_io_args {
1073 /** normal/sendfile/splice */
1074 enum vvp_io_subtype via_io_subtype;
1078 struct kiocb *via_iocb;
1079 struct iov_iter *via_iter;
1082 struct pipe_inode_info *via_pipe;
1083 unsigned int via_flags;
1093 struct ll_cl_context {
1094 struct list_head lcc_list;
1096 const struct lu_env *lcc_env;
1097 struct cl_io *lcc_io;
1098 struct cl_page *lcc_page;
1099 enum lcc_type lcc_type;
1102 struct ll_thread_info {
1103 struct vvp_io_args lti_args;
1104 struct ra_io_arg lti_ria;
1105 struct ll_cl_context lti_io_ctx;
1108 extern struct lu_context_key ll_thread_key;
1110 static inline struct ll_thread_info *ll_env_info(const struct lu_env *env)
1112 struct ll_thread_info *lti;
1114 lti = lu_context_key_get(&env->le_ctx, &ll_thread_key);
1115 LASSERT(lti != NULL);
1120 static inline struct vvp_io_args *ll_env_args(const struct lu_env *env,
1121 enum vvp_io_subtype type)
1123 struct vvp_io_args *via = &ll_env_info(env)->lti_args;
1125 via->via_io_subtype = type;
1130 void ll_io_init(struct cl_io *io, struct file *file, enum cl_io_type iot,
1131 struct vvp_io_args *args);
1133 /* llite/llite_mmap.c */
1135 int ll_teardown_mmaps(struct address_space *mapping, __u64 first, __u64 last);
1136 int ll_file_mmap(struct file * file, struct vm_area_struct * vma);
1137 void policy_from_vma(union ldlm_policy_data *policy, struct vm_area_struct *vma,
1138 unsigned long addr, size_t count);
1139 struct vm_area_struct *our_vma(struct mm_struct *mm, unsigned long addr,
1142 #define ll_s2sbi(sb) (s2lsi(sb)->lsi_llsbi)
1144 /* don't need an addref as the sb_info should be holding one */
1145 static inline struct obd_export *ll_s2dtexp(struct super_block *sb)
1147 return ll_s2sbi(sb)->ll_dt_exp;
1150 /* don't need an addref as the sb_info should be holding one */
1151 static inline struct obd_export *ll_s2mdexp(struct super_block *sb)
1153 return ll_s2sbi(sb)->ll_md_exp;
1156 static inline struct client_obd *sbi2mdc(struct ll_sb_info *sbi)
1158 struct obd_device *obd = sbi->ll_md_exp->exp_obd;
1164 // FIXME: replace the name of this with LL_SB to conform to kernel stuff
1165 static inline struct ll_sb_info *ll_i2sbi(struct inode *inode)
1167 return ll_s2sbi(inode->i_sb);
1170 static inline struct obd_export *ll_i2dtexp(struct inode *inode)
1172 return ll_s2dtexp(inode->i_sb);
1175 static inline struct obd_export *ll_i2mdexp(struct inode *inode)
1177 return ll_s2mdexp(inode->i_sb);
1180 static inline struct lu_fid *ll_inode2fid(struct inode *inode)
1184 LASSERT(inode != NULL);
1185 fid = &ll_i2info(inode)->lli_fid;
1190 static inline bool ll_dir_striped(struct inode *inode)
1193 return S_ISDIR(inode->i_mode) &&
1194 lmv_dir_striped(ll_i2info(inode)->lli_lsm_md);
1197 static inline loff_t ll_file_maxbytes(struct inode *inode)
1199 struct cl_object *obj = ll_i2info(inode)->lli_clob;
1202 return MAX_LFS_FILESIZE;
1204 return min_t(loff_t, cl_object_maxbytes(obj), MAX_LFS_FILESIZE);
1208 extern const struct xattr_handler *ll_xattr_handlers[];
1210 #define XATTR_USER_T 1
1211 #define XATTR_TRUSTED_T 2
1212 #define XATTR_SECURITY_T 3
1213 #define XATTR_ACL_ACCESS_T 4
1214 #define XATTR_ACL_DEFAULT_T 5
1215 #define XATTR_LUSTRE_T 6
1216 #define XATTR_OTHER_T 7
1218 ssize_t ll_listxattr(struct dentry *dentry, char *buffer, size_t size);
1219 int ll_xattr_list(struct inode *inode, const char *name, int type,
1220 void *buffer, size_t size, u64 valid);
1221 const struct xattr_handler *get_xattr_type(const char *name);
1224 * Common IO arguments for various VFS I/O interfaces.
1226 int cl_sb_init(struct super_block *sb);
1227 int cl_sb_fini(struct super_block *sb);
1229 enum ras_update_flags {
1233 void ll_ra_count_put(struct ll_sb_info *sbi, unsigned long len);
1234 void ll_ra_stats_inc(struct inode *inode, enum ra_stat which);
1238 #define LL_SA_RPC_MIN 2
1239 #define LL_SA_RPC_DEF 32
1240 #define LL_SA_RPC_MAX 512
1242 /* XXX: If want to support more concurrent statahead instances,
1243 * please consider to decentralize the RPC lists attached
1244 * on related import, such as imp_{sending,delayed}_list.
1246 #define LL_SA_RUNNING_MAX 256
1247 #define LL_SA_RUNNING_DEF 16
1249 #define LL_SA_CACHE_BIT 5
1250 #define LL_SA_CACHE_SIZE (1 << LL_SA_CACHE_BIT)
1251 #define LL_SA_CACHE_MASK (LL_SA_CACHE_SIZE - 1)
1253 /* per inode struct, for dir only */
1254 struct ll_statahead_info {
1255 struct dentry *sai_dentry;
1256 atomic_t sai_refcount; /* when access this struct, hold
1258 unsigned int sai_max; /* max ahead of lookup */
1259 __u64 sai_sent; /* stat requests sent count */
1260 __u64 sai_replied; /* stat requests which received
1262 __u64 sai_index; /* index of statahead entry */
1263 __u64 sai_index_wait; /* index of entry which is the
1264 * caller is waiting for */
1265 __u64 sai_hit; /* hit count */
1266 __u64 sai_miss; /* miss count:
1267 * for "ls -al" case, includes
1268 * hidden dentry miss;
1269 * for "ls -l" case, it does not
1270 * include hidden dentry miss.
1271 * "sai_miss_hidden" is used for
1274 unsigned int sai_consecutive_miss; /* consecutive miss */
1275 unsigned int sai_miss_hidden;/* "ls -al", but first dentry
1276 * is not a hidden one */
1277 unsigned int sai_skip_hidden;/* skipped hidden dentry count
1279 unsigned int sai_ls_all:1, /* "ls -al", do stat-ahead for
1281 sai_agl_valid:1,/* AGL is valid for the dir */
1282 sai_in_readpage:1;/* statahead is in readdir()*/
1283 wait_queue_head_t sai_waitq; /* stat-ahead wait queue */
1284 struct ptlrpc_thread sai_thread; /* stat-ahead thread */
1285 struct ptlrpc_thread sai_agl_thread; /* AGL thread */
1286 struct list_head sai_interim_entries; /* entries which got async
1287 * stat reply, but not
1289 struct list_head sai_entries; /* completed entries */
1290 struct list_head sai_agls; /* AGLs 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 ll_statahead(struct inode *dir, struct dentry **dentry, bool unplug);
1297 void ll_authorize_statahead(struct inode *dir, void *key);
1298 void ll_deauthorize_statahead(struct inode *dir, void *key);
1301 blkcnt_t dirty_cnt(struct inode *inode);
1303 int cl_glimpse_size0(struct inode *inode, int agl);
1304 int cl_glimpse_lock(const struct lu_env *env, struct cl_io *io,
1305 struct inode *inode, struct cl_object *clob, int agl);
1307 static inline int cl_glimpse_size(struct inode *inode)
1309 return cl_glimpse_size0(inode, 0);
1312 /* AGL is 'asychronous glimpse lock', which is a speculative lock taken as
1313 * part of statahead */
1314 static inline int cl_agl(struct inode *inode)
1316 return cl_glimpse_size0(inode, 1);
1319 int ll_file_lock_ahead(struct file *file, struct llapi_lu_ladvise *ladvise);
1321 int cl_io_get(struct inode *inode, struct lu_env **envout,
1322 struct cl_io **ioout, __u16 *refcheck);
1324 static inline int ll_glimpse_size(struct inode *inode)
1326 struct ll_inode_info *lli = ll_i2info(inode);
1329 down_read(&lli->lli_glimpse_sem);
1330 rc = cl_glimpse_size(inode);
1331 lli->lli_glimpse_time = ktime_get();
1332 up_read(&lli->lli_glimpse_sem);
1336 /* dentry may statahead when statahead is enabled and current process has opened
1337 * parent directory, and this dentry hasn't accessed statahead cache before */
1339 dentry_may_statahead(struct inode *dir, struct dentry *dentry)
1341 struct ll_inode_info *lli;
1342 struct ll_dentry_data *ldd;
1344 if (ll_i2sbi(dir)->ll_sa_max == 0)
1347 lli = ll_i2info(dir);
1349 /* statahead is not allowed for this dir, there may be three causes:
1350 * 1. dir is not opened.
1351 * 2. statahead hit ratio is too low.
1352 * 3. previous stat started statahead thread failed. */
1353 if (!lli->lli_sa_enabled)
1356 /* not the same process, don't statahead */
1357 if (lli->lli_opendir_pid != current_pid())
1361 * When stating a dentry, kernel may trigger 'revalidate' or 'lookup'
1362 * multiple times, eg. for 'getattr', 'getxattr' and etc.
1363 * For patchless client, lookup intent is not accurate, which may
1364 * misguide statahead. For example:
1365 * The 'revalidate' call for 'getattr' and 'getxattr' of a dentry will
1366 * have the same intent -- IT_GETATTR, while one dentry should access
1367 * statahead cache once, otherwise statahead windows is messed up.
1368 * The solution is as following:
1369 * Assign 'lld_sa_generation' with 'lli_sa_generation' when a dentry
1370 * IT_GETATTR for the first time, and subsequent IT_GETATTR will
1371 * bypass interacting with statahead cache by checking
1372 * 'lld_sa_generation == lli->lli_sa_generation'.
1374 ldd = ll_d2d(dentry);
1375 if (ldd != NULL && ldd->lld_sa_generation == lli->lli_sa_generation)
1381 int cl_sync_file_range(struct inode *inode, loff_t start, loff_t end,
1382 enum cl_fsync_mode mode, int ignore_layout);
1384 static inline int ll_file_nolock(const struct file *file)
1386 struct ll_file_data *fd = LUSTRE_FPRIVATE(file);
1387 struct inode *inode = file_inode((struct file *)file);
1389 LASSERT(fd != NULL);
1390 return ((fd->fd_flags & LL_FILE_IGNORE_LOCK) ||
1391 (ll_i2sbi(inode)->ll_flags & LL_SBI_NOLCK));
1394 static inline void ll_set_lock_data(struct obd_export *exp, struct inode *inode,
1395 struct lookup_intent *it, __u64 *bits)
1397 if (!it->it_lock_set) {
1398 struct lustre_handle handle;
1400 /* If this inode is a remote object, it will get two
1401 * separate locks in different namespaces, Master MDT,
1402 * where the name entry is, will grant LOOKUP lock,
1403 * remote MDT, where the object is, will grant
1404 * UPDATE|PERM lock. The inode will be attched to both
1405 * LOOKUP and PERM locks, so revoking either locks will
1406 * case the dcache being cleared */
1407 if (it->it_remote_lock_mode) {
1408 handle.cookie = it->it_remote_lock_handle;
1409 CDEBUG(D_DLMTRACE, "setting l_data to inode "DFID
1410 "(%p) for remote lock %#llx\n",
1411 PFID(ll_inode2fid(inode)), inode,
1413 md_set_lock_data(exp, &handle, inode, NULL);
1416 handle.cookie = it->it_lock_handle;
1418 CDEBUG(D_DLMTRACE, "setting l_data to inode "DFID"(%p)"
1419 " for lock %#llx\n",
1420 PFID(ll_inode2fid(inode)), inode, handle.cookie);
1422 md_set_lock_data(exp, &handle, inode, &it->it_lock_bits);
1423 it->it_lock_set = 1;
1427 *bits = it->it_lock_bits;
1430 static inline int d_lustre_invalid(const struct dentry *dentry)
1432 struct ll_dentry_data *lld = ll_d2d(dentry);
1434 return (lld == NULL) || lld->lld_invalid;
1437 static inline void __d_lustre_invalidate(struct dentry *dentry)
1439 struct ll_dentry_data *lld = ll_d2d(dentry);
1442 lld->lld_invalid = 1;
1446 * Mark dentry INVALID, if dentry refcount is zero (this is normally case for
1447 * ll_md_blocking_ast), unhash this dentry, and let dcache to reclaim it later;
1448 * else dput() of the last refcount will unhash this dentry and kill it.
1450 static inline void d_lustre_invalidate(struct dentry *dentry, int nested)
1452 CDEBUG(D_DENTRY, "invalidate dentry %.*s (%p) parent %p inode %p "
1453 "refc %d\n", dentry->d_name.len, dentry->d_name.name, dentry,
1454 dentry->d_parent, dentry->d_inode, ll_d_count(dentry));
1456 spin_lock_nested(&dentry->d_lock,
1457 nested ? DENTRY_D_LOCK_NESTED : DENTRY_D_LOCK_NORMAL);
1458 __d_lustre_invalidate(dentry);
1460 * We should be careful about dentries created by d_obtain_alias().
1461 * These dentries are not put in the dentry tree, instead they are
1462 * linked to sb->s_anon through dentry->d_hash.
1463 * shrink_dcache_for_umount() shrinks the tree and sb->s_anon list.
1464 * If we unhashed such a dentry, unmount would not be able to find
1465 * it and busy inodes would be reported.
1467 if (ll_d_count(dentry) == 0 && !(dentry->d_flags & DCACHE_DISCONNECTED))
1469 spin_unlock(&dentry->d_lock);
1472 static inline void d_lustre_revalidate(struct dentry *dentry)
1474 spin_lock(&dentry->d_lock);
1475 LASSERT(ll_d2d(dentry) != NULL);
1476 ll_d2d(dentry)->lld_invalid = 0;
1477 spin_unlock(&dentry->d_lock);
1480 static inline dev_t ll_compat_encode_dev(dev_t dev)
1482 /* The compat_sys_*stat*() syscalls will fail unless the
1483 * device majors and minors are both less than 256. Note that
1484 * the value returned here will be passed through
1485 * old_encode_dev() in cp_compat_stat(). And so we are not
1486 * trying to return a valid compat (u16) device number, just
1487 * one that will pass the old_valid_dev() check. */
1489 return MKDEV(MAJOR(dev) & 0xff, MINOR(dev) & 0xff);
1492 int ll_layout_conf(struct inode *inode, const struct cl_object_conf *conf);
1493 int ll_layout_refresh(struct inode *inode, __u32 *gen);
1494 int ll_layout_restore(struct inode *inode, loff_t start, __u64 length);
1495 int ll_layout_write_intent(struct inode *inode, enum layout_intent_opc opc,
1496 struct lu_extent *ext);
1498 int ll_xattr_init(void);
1499 void ll_xattr_fini(void);
1501 int ll_page_sync_io(const struct lu_env *env, struct cl_io *io,
1502 struct cl_page *page, enum cl_req_type crt);
1504 int ll_getparent(struct file *file, struct getparent __user *arg);
1507 int cl_setattr_ost(struct cl_object *obj, const struct iattr *attr,
1508 enum op_xvalid xvalid, unsigned int attr_flags);
1510 extern struct lu_env *cl_inode_fini_env;
1511 extern __u16 cl_inode_fini_refcheck;
1513 int cl_file_inode_init(struct inode *inode, struct lustre_md *md);
1514 void cl_inode_fini(struct inode *inode);
1516 u64 cl_fid_build_ino(const struct lu_fid *fid, int api32);
1517 u32 cl_fid_build_gen(const struct lu_fid *fid);
1519 static inline struct pcc_super *ll_i2pccs(struct inode *inode)
1521 return &ll_i2sbi(inode)->ll_pcc_super;
1524 static inline struct pcc_super *ll_info2pccs(struct ll_inode_info *lli)
1526 return ll_i2pccs(ll_info2i(lli));
1529 #endif /* LLITE_INTERNAL_H */