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 /* serialize normal readdir and statahead-readdir. */
147 struct mutex lli_readdir_mutex;
149 /* metadata statahead */
150 /* since parent-child threads can share the same @file
151 * struct, "opendir_key" is the token when dir close for
152 * case of parent exit before child -- it is me should
153 * cleanup the dir readahead. */
154 void *lli_opendir_key;
155 struct ll_statahead_info *lli_sai;
156 /* protect statahead stuff. */
157 spinlock_t lli_sa_lock;
158 /* "opendir_pid" is the token when lookup/revalid
159 * -- I am the owner of dir statahead. */
160 pid_t lli_opendir_pid;
161 /* stat will try to access statahead entries or start
162 * statahead if this flag is set, and this flag will be
163 * set upon dir open, and cleared when dir is closed,
164 * statahead hit ratio is too low, or start statahead
166 unsigned int lli_sa_enabled:1;
167 /* generation for statahead */
168 unsigned int lli_sa_generation;
169 /* rw lock protects lli_lsm_md */
170 struct rw_semaphore lli_lsm_sem;
171 /* directory stripe information */
172 struct lmv_stripe_md *lli_lsm_md;
173 /* directory default LMV */
174 struct lmv_stripe_md *lli_default_lsm_md;
177 /* for non-directory */
179 struct mutex lli_size_mutex;
180 char *lli_symlink_name;
182 * struct rw_semaphore {
183 * signed long count; // align d.d_def_acl
184 * spinlock_t wait_lock; // align d.d_sa_lock
185 * struct list_head wait_list;
188 struct rw_semaphore lli_trunc_sem;
189 struct range_lock_tree lli_write_tree;
191 struct rw_semaphore lli_glimpse_sem;
192 ktime_t lli_glimpse_time;
193 struct list_head lli_agl_list;
196 /* for writepage() only to communicate to fsync */
199 /* protect the file heat fields */
200 spinlock_t lli_heat_lock;
201 __u32 lli_heat_flags;
202 struct obd_heat_instance lli_heat_instances[OBD_HEAT_COUNT];
205 * Whenever a process try to read/write the file, the
206 * jobid of the process will be saved here, and it'll
207 * be packed into the write PRC when flush later.
209 * So the read/write statistics for jobid will not be
210 * accurate if the file is shared by different jobs.
212 char lli_jobid[LUSTRE_JOBID_SIZE];
214 struct mutex lli_pcc_lock;
215 enum lu_pcc_state_flags lli_pcc_state;
216 struct pcc_inode *lli_pcc_inode;
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 unsigned long ria_start; /* start offset of read-ahead*/
404 unsigned long ria_end; /* end offset of read-ahead*/
405 unsigned long ria_reserved; /* reserved pages for read-ahead */
406 unsigned long 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*/
412 /* ria_length and ria_pages 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_pages;
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_umounting:1,
543 ll_xattr_cache_enabled:1,
544 ll_xattr_cache_set:1, /* already set to 0/1 */
545 ll_client_common_fill_super_succeeded:1,
548 struct lustre_client_ocd ll_lco;
550 struct lprocfs_stats *ll_stats; /* lprocfs stats counter */
552 /* Used to track "unstable" pages on a client, and maintain a
553 * LRU list of clean pages. An "unstable" page is defined as
554 * any page which is sent to a server as part of a bulk request,
555 * but is uncommitted to stable storage. */
556 struct cl_client_cache *ll_cache;
558 struct lprocfs_stats *ll_ra_stats;
560 struct ll_ra_info ll_ra_info;
561 unsigned int ll_namelen;
562 struct file_operations *ll_fop;
564 struct lu_site *ll_site;
565 struct cl_device *ll_cl;
567 struct ll_rw_extents_info ll_rw_extents_info;
568 int ll_extent_process_count;
569 struct ll_rw_process_info ll_rw_process_info[LL_PROCESS_HIST_MAX];
570 unsigned int ll_offset_process_count;
571 struct ll_rw_process_info ll_rw_offset_info[LL_OFFSET_HIST_MAX];
572 unsigned int ll_rw_offset_entry_count;
573 int ll_stats_track_id;
574 enum stats_track_type ll_stats_track_type;
577 /* metadata stat-ahead */
578 unsigned int ll_sa_running_max;/* max concurrent
579 * statahead instances */
580 unsigned int ll_sa_max; /* max statahead RPCs */
581 atomic_t ll_sa_total; /* statahead thread started
583 atomic_t ll_sa_wrong; /* statahead thread stopped for
585 atomic_t ll_sa_running; /* running statahead thread
587 atomic_t ll_agl_total; /* AGL thread started count */
589 dev_t ll_sdev_orig; /* save s_dev before assign for
592 struct root_squash_info ll_squash;
595 /* st_blksize returned by stat(2), when non-zero */
596 unsigned int ll_stat_blksize;
598 /* maximum relative age of cached statfs results */
599 unsigned int ll_statfs_max_age;
601 struct kset ll_kset; /* sysfs object */
602 struct completion ll_kobj_unregister;
605 unsigned int ll_heat_decay_weight;
606 unsigned int ll_heat_period_second;
608 /* filesystem fsname */
609 char ll_fsname[LUSTRE_MAXFSNAME + 1];
611 /* Persistent Client Cache */
612 struct pcc_super ll_pcc_super;
615 #define SBI_DEFAULT_HEAT_DECAY_WEIGHT ((80 * 256 + 50) / 100)
616 #define SBI_DEFAULT_HEAT_PERIOD_SECOND (60)
618 * per file-descriptor read-ahead data.
620 struct ll_readahead_state {
623 * index of the last page that read(2) needed and that wasn't in the
624 * cache. Used by ras_update() to detect seeks.
626 * XXX nikita: if access seeks into cached region, Lustre doesn't see
629 unsigned long ras_last_readpage;
631 * number of pages read after last read-ahead window reset. As window
632 * is reset on each seek, this is effectively a number of consecutive
633 * accesses. Maybe ->ras_accessed_in_window is better name.
635 * XXX nikita: window is also reset (by ras_update()) when Lustre
636 * believes that memory pressure evicts read-ahead pages. In that
637 * case, it probably doesn't make sense to expand window to
638 * PTLRPC_MAX_BRW_PAGES on the third access.
640 unsigned long ras_consecutive_pages;
642 * number of read requests after the last read-ahead window reset
643 * As window is reset on each seek, this is effectively the number
644 * on consecutive read request and is used to trigger read-ahead.
646 unsigned long ras_consecutive_requests;
648 * Parameters of current read-ahead window. Handled by
649 * ras_update(). On the initial access to the file or after a seek,
650 * window is reset to 0. After 3 consecutive accesses, window is
651 * expanded to PTLRPC_MAX_BRW_PAGES. Afterwards, window is enlarged by
652 * PTLRPC_MAX_BRW_PAGES chunks up to ->ra_max_pages.
654 unsigned long ras_window_start, ras_window_len;
656 * Optimal RPC size. It decides how many pages will be sent
657 * for each read-ahead.
659 unsigned long ras_rpc_size;
661 * Where next read-ahead should start at. This lies within read-ahead
662 * window. Read-ahead window is read in pieces rather than at once
663 * because: 1. lustre limits total number of pages under read-ahead by
664 * ->ra_max_pages (see ll_ra_count_get()), 2. client cannot read pages
665 * not covered by DLM lock.
667 unsigned long ras_next_readahead;
669 * Total number of ll_file_read requests issued, reads originating
670 * due to mmap are not counted in this total. This value is used to
671 * trigger full file read-ahead after multiple reads to a small file.
673 unsigned long ras_requests;
675 * Page index with respect to the current request, these value
676 * will not be accurate when dealing with reads issued via mmap.
678 unsigned long ras_request_index;
680 * The following 3 items are used for detecting the stride I/O
682 * In stride I/O mode,
683 * ...............|-----data-----|****gap*****|--------|******|....
684 * offset |-stride_pages-|-stride_gap-|
685 * ras_stride_offset = offset;
686 * ras_stride_length = stride_pages + stride_gap;
687 * ras_stride_pages = stride_pages;
688 * Note: all these three items are counted by pages.
690 unsigned long ras_stride_length;
691 unsigned long ras_stride_pages;
692 pgoff_t ras_stride_offset;
694 * number of consecutive stride request count, and it is similar as
695 * ras_consecutive_requests, but used for stride I/O mode.
696 * Note: only more than 2 consecutive stride request are detected,
697 * stride read-ahead will be enable
699 unsigned long ras_consecutive_stride_requests;
700 /* index of the last page that async readahead starts */
701 unsigned long ras_async_last_readpage;
704 struct ll_readahead_work {
705 /** File to readahead */
706 struct file *lrw_file;
708 unsigned long lrw_start;
710 unsigned long lrw_end;
712 /* async worker to handler read */
713 struct work_struct lrw_readahead_work;
716 extern struct kmem_cache *ll_file_data_slab;
717 struct lustre_handle;
718 struct ll_file_data {
719 struct ll_readahead_state fd_ras;
720 struct ll_grouplock fd_grouplock;
724 /* openhandle if lease exists for this file.
725 * Borrow lli->lli_och_mutex to protect assignment */
726 struct obd_client_handle *fd_lease_och;
727 struct obd_client_handle *fd_och;
728 struct file *fd_file;
729 /* Indicate whether need to report failure when close.
730 * true: failure is known, not report again.
731 * false: unknown failure, should report. */
732 bool fd_write_failed;
733 bool ll_lock_no_expand;
734 rwlock_t fd_lock; /* protect lcc list */
735 struct list_head fd_lccs; /* list of ll_cl_context */
736 /* Used by mirrored file to lead IOs to a specific mirror, usually
737 * for mirror resync. 0 means default. */
738 __u32 fd_designated_mirror;
739 /* The layout version when resync starts. Resync I/O should carry this
740 * layout version for verification to OST objects */
741 __u32 fd_layout_version;
742 struct pcc_file fd_pcc_file;
745 void llite_tunables_unregister(void);
746 int llite_tunables_register(void);
748 static inline struct inode *ll_info2i(struct ll_inode_info *lli)
750 return &lli->lli_vfs_inode;
753 __u32 ll_i2suppgid(struct inode *i);
754 void ll_i2gids(__u32 *suppgids, struct inode *i1,struct inode *i2);
756 static inline int ll_need_32bit_api(struct ll_sb_info *sbi)
758 #if BITS_PER_LONG == 32
760 #elif defined(CONFIG_COMPAT)
761 if (unlikely(sbi->ll_flags & LL_SBI_32BIT_API))
764 # ifdef CONFIG_X86_X32
765 /* in_compat_syscall() returns true when called from a kthread
766 * and CONFIG_X86_X32 is enabled, which is wrong. So check
767 * whether the caller comes from a syscall (ie. not a kthread)
768 * before calling in_compat_syscall(). */
769 if (current->flags & PF_KTHREAD)
773 return unlikely(in_compat_syscall());
775 return unlikely(sbi->ll_flags & LL_SBI_32BIT_API);
779 static inline bool ll_sbi_has_fast_read(struct ll_sb_info *sbi)
781 return !!(sbi->ll_flags & LL_SBI_FAST_READ);
784 static inline bool ll_sbi_has_tiny_write(struct ll_sb_info *sbi)
786 return !!(sbi->ll_flags & LL_SBI_TINY_WRITE);
789 static inline bool ll_sbi_has_file_heat(struct ll_sb_info *sbi)
791 return !!(sbi->ll_flags & LL_SBI_FILE_HEAT);
794 void ll_ras_enter(struct file *f);
796 /* llite/lcommon_misc.c */
797 int cl_ocd_update(struct obd_device *host, struct obd_device *watched,
798 enum obd_notify_event ev, void *owner);
799 int cl_get_grouplock(struct cl_object *obj, unsigned long gid, int nonblock,
800 struct ll_grouplock *lg);
801 void cl_put_grouplock(struct ll_grouplock *lg);
803 /* llite/lproc_llite.c */
804 int ll_debugfs_register_super(struct super_block *sb, const char *name);
805 void ll_debugfs_unregister_super(struct super_block *sb);
806 void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count);
810 LPROC_LL_DIRTY_MISSES,
812 LPROC_LL_WRITE_BYTES,
837 LPROC_LL_ALLOC_INODE,
840 LPROC_LL_GETXATTR_HITS,
842 LPROC_LL_REMOVEXATTR,
844 LPROC_LL_FILE_OPCODES
848 struct ll_dir_chain {
851 static inline void ll_dir_chain_init(struct ll_dir_chain *chain)
855 static inline void ll_dir_chain_fini(struct ll_dir_chain *chain)
859 extern const struct file_operations ll_dir_operations;
860 extern const struct inode_operations ll_dir_inode_operations;
861 #ifdef HAVE_DIR_CONTEXT
862 int ll_dir_read(struct inode *inode, __u64 *pos, struct md_op_data *op_data,
863 struct dir_context *ctx);
865 int ll_dir_read(struct inode *inode, __u64 *pos, struct md_op_data *op_data,
866 void *cookie, filldir_t filldir);
868 int ll_get_mdt_idx(struct inode *inode);
869 int ll_get_mdt_idx_by_fid(struct ll_sb_info *sbi, const struct lu_fid *fid);
870 struct page *ll_get_dir_page(struct inode *dir, struct md_op_data *op_data,
871 __u64 offset, struct ll_dir_chain *chain);
872 void ll_release_page(struct inode *inode, struct page *page, bool remove);
875 extern const struct inode_operations ll_special_inode_operations;
877 struct inode *ll_iget(struct super_block *sb, ino_t hash,
878 struct lustre_md *lic);
879 int ll_test_inode_by_fid(struct inode *inode, void *opaque);
880 int ll_md_blocking_ast(struct ldlm_lock *, struct ldlm_lock_desc *,
881 void *data, int flag);
882 struct dentry *ll_splice_alias(struct inode *inode, struct dentry *de);
883 int ll_rmdir_entry(struct inode *dir, char *name, int namelen);
884 void ll_update_times(struct ptlrpc_request *request, struct inode *inode);
887 int ll_writepage(struct page *page, struct writeback_control *wbc);
888 int ll_writepages(struct address_space *, struct writeback_control *wbc);
889 int ll_readpage(struct file *file, struct page *page);
890 int ll_io_read_page(const struct lu_env *env, struct cl_io *io,
891 struct cl_page *page, struct file *file);
892 void ll_readahead_init(struct inode *inode, struct ll_readahead_state *ras);
893 int vvp_io_write_commit(const struct lu_env *env, struct cl_io *io);
896 void ll_cl_add(struct file *file, const struct lu_env *env, struct cl_io *io,
898 void ll_cl_remove(struct file *file, const struct lu_env *env);
899 struct ll_cl_context *ll_cl_find(struct file *file);
901 extern const struct address_space_operations ll_aops;
904 extern struct file_operations ll_file_operations;
905 extern struct file_operations ll_file_operations_flock;
906 extern struct file_operations ll_file_operations_noflock;
907 extern struct inode_operations ll_file_inode_operations;
908 extern int ll_have_md_lock(struct inode *inode, __u64 *bits,
909 enum ldlm_mode l_req_mode);
910 extern enum ldlm_mode ll_take_md_lock(struct inode *inode, __u64 bits,
911 struct lustre_handle *lockh, __u64 flags,
912 enum ldlm_mode mode);
914 int ll_file_open(struct inode *inode, struct file *file);
915 int ll_file_release(struct inode *inode, struct file *file);
916 int ll_release_openhandle(struct dentry *, struct lookup_intent *);
917 int ll_md_real_close(struct inode *inode, fmode_t fmode);
918 extern void ll_rw_stats_tally(struct ll_sb_info *sbi, pid_t pid,
919 struct ll_file_data *file, loff_t pos,
920 size_t count, int rw);
921 void ll_io_init(struct cl_io *io, struct file *file, enum cl_io_type iot);
922 #ifdef HAVE_INODEOPS_ENHANCED_GETATTR
923 int ll_getattr(const struct path *path, struct kstat *stat,
924 u32 request_mask, unsigned int flags);
926 int ll_getattr(struct vfsmount *mnt, struct dentry *de, struct kstat *stat);
928 int ll_getattr_dentry(struct dentry *de, struct kstat *stat);
929 struct posix_acl *ll_get_acl(struct inode *inode, int type);
930 #ifdef HAVE_IOP_SET_ACL
931 #ifdef CONFIG_FS_POSIX_ACL
932 int ll_set_acl(struct inode *inode, struct posix_acl *acl, int type);
933 #else /* !CONFIG_FS_POSIX_ACL */
934 #define ll_set_acl NULL
935 #endif /* CONFIG_FS_POSIX_ACL */
938 int ll_migrate(struct inode *parent, struct file *file,
939 struct lmv_user_md *lum, const char *name);
940 int ll_get_fid_by_name(struct inode *parent, const char *name,
941 int namelen, struct lu_fid *fid, struct inode **inode);
942 #ifdef HAVE_GENERIC_PERMISSION_4ARGS
943 int ll_inode_permission(struct inode *inode, int mask, unsigned int flags);
945 # ifndef HAVE_INODE_PERMISION_2ARGS
946 int ll_inode_permission(struct inode *inode, int mask, struct nameidata *nd);
948 int ll_inode_permission(struct inode *inode, int mask);
951 int ll_ioctl_check_project(struct inode *inode, struct fsxattr *fa);
952 int ll_ioctl_fsgetxattr(struct inode *inode, unsigned int cmd,
954 int ll_ioctl_fssetxattr(struct inode *inode, unsigned int cmd,
957 int ll_lov_setstripe_ea_info(struct inode *inode, struct dentry *dentry,
958 __u64 flags, struct lov_user_md *lum,
960 int ll_lov_getstripe_ea_info(struct inode *inode, const char *filename,
961 struct lov_mds_md **lmm, int *lmm_size,
962 struct ptlrpc_request **request);
963 int ll_dir_setstripe(struct inode *inode, struct lov_user_md *lump,
965 int ll_dir_getstripe(struct inode *inode, void **lmmp,
966 int *lmm_size, struct ptlrpc_request **request,
968 int ll_fsync(struct file *file, loff_t start, loff_t end, int data);
969 int ll_merge_attr(const struct lu_env *env, struct inode *inode);
970 int ll_fid2path(struct inode *inode, void __user *arg);
971 int ll_data_version(struct inode *inode, __u64 *data_version, int flags);
972 int ll_hsm_release(struct inode *inode);
973 int ll_hsm_state_set(struct inode *inode, struct hsm_state_set *hss);
974 void ll_io_set_mirror(struct cl_io *io, const struct file *file);
978 int ll_d_init(struct dentry *de);
979 extern const struct dentry_operations ll_d_ops;
980 void ll_intent_drop_lock(struct lookup_intent *);
981 void ll_intent_release(struct lookup_intent *);
982 void ll_invalidate_aliases(struct inode *);
983 void ll_lookup_finish_locks(struct lookup_intent *it, struct dentry *dentry);
984 int ll_revalidate_it_finish(struct ptlrpc_request *request,
985 struct lookup_intent *it, struct dentry *de);
987 /* llite/llite_lib.c */
988 extern struct super_operations lustre_super_operations;
990 void ll_lli_init(struct ll_inode_info *lli);
991 int ll_fill_super(struct super_block *sb, struct vfsmount *mnt);
992 void ll_put_super(struct super_block *sb);
993 void ll_kill_super(struct super_block *sb);
994 struct inode *ll_inode_from_resource_lock(struct ldlm_lock *lock);
995 void ll_dir_clear_lsm_md(struct inode *inode);
996 void ll_clear_inode(struct inode *inode);
997 int ll_setattr_raw(struct dentry *dentry, struct iattr *attr,
998 enum op_xvalid xvalid, bool hsm_import);
999 int ll_setattr(struct dentry *de, struct iattr *attr);
1000 int ll_statfs(struct dentry *de, struct kstatfs *sfs);
1001 int ll_statfs_internal(struct ll_sb_info *sbi, struct obd_statfs *osfs,
1003 int ll_update_inode(struct inode *inode, struct lustre_md *md);
1004 void ll_update_inode_flags(struct inode *inode, int ext_flags);
1005 int ll_read_inode2(struct inode *inode, void *opaque);
1006 void ll_delete_inode(struct inode *inode);
1007 int ll_iocontrol(struct inode *inode, struct file *file,
1008 unsigned int cmd, unsigned long arg);
1009 int ll_flush_ctx(struct inode *inode);
1010 void ll_umount_begin(struct super_block *sb);
1011 int ll_remount_fs(struct super_block *sb, int *flags, char *data);
1012 #ifdef HAVE_SUPEROPS_USE_DENTRY
1013 int ll_show_options(struct seq_file *seq, struct dentry *dentry);
1015 int ll_show_options(struct seq_file *seq, struct vfsmount *vfs);
1017 void ll_dirty_page_discard_warn(struct page *page, int ioret);
1018 int ll_prep_inode(struct inode **inode, struct ptlrpc_request *req,
1019 struct super_block *, struct lookup_intent *);
1020 int ll_obd_statfs(struct inode *inode, void __user *arg);
1021 int ll_get_max_mdsize(struct ll_sb_info *sbi, int *max_mdsize);
1022 int ll_get_default_mdsize(struct ll_sb_info *sbi, int *default_mdsize);
1023 int ll_set_default_mdsize(struct ll_sb_info *sbi, int default_mdsize);
1025 void ll_unlock_md_op_lsm(struct md_op_data *op_data);
1026 struct md_op_data *ll_prep_md_op_data(struct md_op_data *op_data,
1027 struct inode *i1, struct inode *i2,
1028 const char *name, size_t namelen,
1029 __u32 mode, enum md_op_code opc,
1031 void ll_finish_md_op_data(struct md_op_data *op_data);
1032 int ll_get_obd_name(struct inode *inode, unsigned int cmd, unsigned long arg);
1033 void ll_compute_rootsquash_state(struct ll_sb_info *sbi);
1034 ssize_t ll_copy_user_md(const struct lov_user_md __user *md,
1035 struct lov_user_md **kbuf);
1036 void ll_open_cleanup(struct super_block *sb, struct ptlrpc_request *open_req);
1038 void ll_dom_finish_open(struct inode *inode, struct ptlrpc_request *req,
1039 struct lookup_intent *it);
1041 /* Compute expected user md size when passing in a md from user space */
1042 static inline ssize_t ll_lov_user_md_size(const struct lov_user_md *lum)
1044 switch (lum->lmm_magic) {
1045 case LOV_USER_MAGIC_V1:
1046 return sizeof(struct lov_user_md_v1);
1047 case LOV_USER_MAGIC_V3:
1048 return sizeof(struct lov_user_md_v3);
1049 case LOV_USER_MAGIC_SPECIFIC:
1050 if (lum->lmm_stripe_count > LOV_MAX_STRIPE_COUNT)
1053 return lov_user_md_size(lum->lmm_stripe_count,
1054 LOV_USER_MAGIC_SPECIFIC);
1055 case LOV_USER_MAGIC_COMP_V1:
1056 return ((struct lov_comp_md_v1 *)lum)->lcm_size;
1057 case LOV_USER_MAGIC_FOREIGN:
1058 return foreign_size(lum);
1064 /* llite/llite_nfs.c */
1065 extern struct export_operations lustre_export_operations;
1066 __u32 get_uuid2int(const char *name, int len);
1067 struct inode *search_inode_for_lustre(struct super_block *sb,
1068 const struct lu_fid *fid);
1069 int ll_dir_get_parent_fid(struct inode *dir, struct lu_fid *parent_fid);
1071 /* llite/symlink.c */
1072 extern struct inode_operations ll_fast_symlink_inode_operations;
1075 * IO arguments for various VFS I/O interfaces.
1077 struct vvp_io_args {
1078 /** normal/sendfile/splice */
1079 enum vvp_io_subtype via_io_subtype;
1083 struct kiocb *via_iocb;
1084 struct iov_iter *via_iter;
1087 struct pipe_inode_info *via_pipe;
1088 unsigned int via_flags;
1098 struct ll_cl_context {
1099 struct list_head lcc_list;
1101 const struct lu_env *lcc_env;
1102 struct cl_io *lcc_io;
1103 struct cl_page *lcc_page;
1104 enum lcc_type lcc_type;
1107 struct ll_thread_info {
1108 struct iov_iter lti_iter;
1109 struct vvp_io_args lti_args;
1110 struct ra_io_arg lti_ria;
1111 struct ll_cl_context lti_io_ctx;
1114 extern struct lu_context_key ll_thread_key;
1116 static inline struct ll_thread_info *ll_env_info(const struct lu_env *env)
1118 struct ll_thread_info *lti;
1120 lti = lu_context_key_get(&env->le_ctx, &ll_thread_key);
1121 LASSERT(lti != NULL);
1126 static inline struct vvp_io_args *ll_env_args(const struct lu_env *env,
1127 enum vvp_io_subtype type)
1129 struct vvp_io_args *via = &ll_env_info(env)->lti_args;
1131 via->via_io_subtype = type;
1136 /* llite/llite_mmap.c */
1138 int ll_teardown_mmaps(struct address_space *mapping, __u64 first, __u64 last);
1139 int ll_file_mmap(struct file * file, struct vm_area_struct * vma);
1140 void policy_from_vma(union ldlm_policy_data *policy, struct vm_area_struct *vma,
1141 unsigned long addr, size_t count);
1142 struct vm_area_struct *our_vma(struct mm_struct *mm, unsigned long addr,
1145 static inline void ll_invalidate_page(struct page *vmpage)
1147 struct address_space *mapping = vmpage->mapping;
1148 loff_t offset = vmpage->index << PAGE_SHIFT;
1150 LASSERT(PageLocked(vmpage));
1151 if (mapping == NULL)
1155 * truncate_complete_page() calls
1156 * a_ops->invalidatepage()->cl_page_delete()->vvp_page_delete().
1158 ll_teardown_mmaps(mapping, offset, offset + PAGE_SIZE);
1159 truncate_complete_page(mapping, vmpage);
1162 #define ll_s2sbi(sb) (s2lsi(sb)->lsi_llsbi)
1164 /* don't need an addref as the sb_info should be holding one */
1165 static inline struct obd_export *ll_s2dtexp(struct super_block *sb)
1167 return ll_s2sbi(sb)->ll_dt_exp;
1170 /* don't need an addref as the sb_info should be holding one */
1171 static inline struct obd_export *ll_s2mdexp(struct super_block *sb)
1173 return ll_s2sbi(sb)->ll_md_exp;
1176 static inline struct client_obd *sbi2mdc(struct ll_sb_info *sbi)
1178 struct obd_device *obd = sbi->ll_md_exp->exp_obd;
1184 // FIXME: replace the name of this with LL_SB to conform to kernel stuff
1185 static inline struct ll_sb_info *ll_i2sbi(struct inode *inode)
1187 return ll_s2sbi(inode->i_sb);
1190 static inline struct obd_export *ll_i2dtexp(struct inode *inode)
1192 return ll_s2dtexp(inode->i_sb);
1195 static inline struct obd_export *ll_i2mdexp(struct inode *inode)
1197 return ll_s2mdexp(inode->i_sb);
1200 static inline struct lu_fid *ll_inode2fid(struct inode *inode)
1204 LASSERT(inode != NULL);
1205 fid = &ll_i2info(inode)->lli_fid;
1210 static inline bool ll_dir_striped(struct inode *inode)
1213 return S_ISDIR(inode->i_mode) &&
1214 lmv_dir_striped(ll_i2info(inode)->lli_lsm_md);
1217 static inline loff_t ll_file_maxbytes(struct inode *inode)
1219 struct cl_object *obj = ll_i2info(inode)->lli_clob;
1222 return MAX_LFS_FILESIZE;
1224 return min_t(loff_t, cl_object_maxbytes(obj), MAX_LFS_FILESIZE);
1228 extern const struct xattr_handler *ll_xattr_handlers[];
1230 #define XATTR_USER_T 1
1231 #define XATTR_TRUSTED_T 2
1232 #define XATTR_SECURITY_T 3
1233 #define XATTR_ACL_ACCESS_T 4
1234 #define XATTR_ACL_DEFAULT_T 5
1235 #define XATTR_LUSTRE_T 6
1236 #define XATTR_OTHER_T 7
1238 ssize_t ll_listxattr(struct dentry *dentry, char *buffer, size_t size);
1239 int ll_xattr_list(struct inode *inode, const char *name, int type,
1240 void *buffer, size_t size, u64 valid);
1241 const struct xattr_handler *get_xattr_type(const char *name);
1244 * Common IO arguments for various VFS I/O interfaces.
1246 int cl_sb_init(struct super_block *sb);
1247 int cl_sb_fini(struct super_block *sb);
1249 enum ras_update_flags {
1253 void ll_ra_count_put(struct ll_sb_info *sbi, unsigned long len);
1254 void ll_ra_stats_inc(struct inode *inode, enum ra_stat which);
1258 #define LL_SA_RPC_MIN 2
1259 #define LL_SA_RPC_DEF 32
1260 #define LL_SA_RPC_MAX 512
1262 /* XXX: If want to support more concurrent statahead instances,
1263 * please consider to decentralize the RPC lists attached
1264 * on related import, such as imp_{sending,delayed}_list.
1266 #define LL_SA_RUNNING_MAX 256
1267 #define LL_SA_RUNNING_DEF 16
1269 #define LL_SA_CACHE_BIT 5
1270 #define LL_SA_CACHE_SIZE (1 << LL_SA_CACHE_BIT)
1271 #define LL_SA_CACHE_MASK (LL_SA_CACHE_SIZE - 1)
1273 /* per inode struct, for dir only */
1274 struct ll_statahead_info {
1275 struct dentry *sai_dentry;
1276 atomic_t sai_refcount; /* when access this struct, hold
1278 unsigned int sai_max; /* max ahead of lookup */
1279 __u64 sai_sent; /* stat requests sent count */
1280 __u64 sai_replied; /* stat requests which received
1282 __u64 sai_index; /* index of statahead entry */
1283 __u64 sai_index_wait; /* index of entry which is the
1284 * caller is waiting for */
1285 __u64 sai_hit; /* hit count */
1286 __u64 sai_miss; /* miss count:
1287 * for "ls -al" case, includes
1288 * hidden dentry miss;
1289 * for "ls -l" case, it does not
1290 * include hidden dentry miss.
1291 * "sai_miss_hidden" is used for
1294 unsigned int sai_consecutive_miss; /* consecutive miss */
1295 unsigned int sai_miss_hidden;/* "ls -al", but first dentry
1296 * is not a hidden one */
1297 unsigned int sai_skip_hidden;/* skipped hidden dentry count
1299 unsigned int sai_ls_all:1, /* "ls -al", do stat-ahead for
1301 sai_agl_valid:1,/* AGL is valid for the dir */
1302 sai_in_readpage:1;/* statahead is in readdir()*/
1303 wait_queue_head_t sai_waitq; /* stat-ahead wait queue */
1304 struct ptlrpc_thread sai_thread; /* stat-ahead thread */
1305 struct ptlrpc_thread sai_agl_thread; /* AGL thread */
1306 struct list_head sai_interim_entries; /* entries which got async
1307 * stat reply, but not
1309 struct list_head sai_entries; /* completed entries */
1310 struct list_head sai_agls; /* AGLs to be sent */
1311 struct list_head sai_cache[LL_SA_CACHE_SIZE];
1312 spinlock_t sai_cache_lock[LL_SA_CACHE_SIZE];
1313 atomic_t sai_cache_count; /* entry count in cache */
1316 int ll_statahead(struct inode *dir, struct dentry **dentry, bool unplug);
1317 void ll_authorize_statahead(struct inode *dir, void *key);
1318 void ll_deauthorize_statahead(struct inode *dir, void *key);
1321 blkcnt_t dirty_cnt(struct inode *inode);
1323 int cl_glimpse_size0(struct inode *inode, int agl);
1324 int cl_glimpse_lock(const struct lu_env *env, struct cl_io *io,
1325 struct inode *inode, struct cl_object *clob, int agl);
1327 static inline int cl_glimpse_size(struct inode *inode)
1329 return cl_glimpse_size0(inode, 0);
1332 /* AGL is 'asychronous glimpse lock', which is a speculative lock taken as
1333 * part of statahead */
1334 static inline int cl_agl(struct inode *inode)
1336 return cl_glimpse_size0(inode, 1);
1339 int ll_file_lock_ahead(struct file *file, struct llapi_lu_ladvise *ladvise);
1341 int cl_io_get(struct inode *inode, struct lu_env **envout,
1342 struct cl_io **ioout, __u16 *refcheck);
1344 static inline int ll_glimpse_size(struct inode *inode)
1346 struct ll_inode_info *lli = ll_i2info(inode);
1349 down_read(&lli->lli_glimpse_sem);
1350 rc = cl_glimpse_size(inode);
1351 lli->lli_glimpse_time = ktime_get();
1352 up_read(&lli->lli_glimpse_sem);
1356 /* dentry may statahead when statahead is enabled and current process has opened
1357 * parent directory, and this dentry hasn't accessed statahead cache before */
1359 dentry_may_statahead(struct inode *dir, struct dentry *dentry)
1361 struct ll_inode_info *lli;
1362 struct ll_dentry_data *ldd;
1364 if (ll_i2sbi(dir)->ll_sa_max == 0)
1367 lli = ll_i2info(dir);
1369 /* statahead is not allowed for this dir, there may be three causes:
1370 * 1. dir is not opened.
1371 * 2. statahead hit ratio is too low.
1372 * 3. previous stat started statahead thread failed. */
1373 if (!lli->lli_sa_enabled)
1376 /* not the same process, don't statahead */
1377 if (lli->lli_opendir_pid != current_pid())
1381 * When stating a dentry, kernel may trigger 'revalidate' or 'lookup'
1382 * multiple times, eg. for 'getattr', 'getxattr' and etc.
1383 * For patchless client, lookup intent is not accurate, which may
1384 * misguide statahead. For example:
1385 * The 'revalidate' call for 'getattr' and 'getxattr' of a dentry will
1386 * have the same intent -- IT_GETATTR, while one dentry should access
1387 * statahead cache once, otherwise statahead windows is messed up.
1388 * The solution is as following:
1389 * Assign 'lld_sa_generation' with 'lli_sa_generation' when a dentry
1390 * IT_GETATTR for the first time, and subsequent IT_GETATTR will
1391 * bypass interacting with statahead cache by checking
1392 * 'lld_sa_generation == lli->lli_sa_generation'.
1394 ldd = ll_d2d(dentry);
1395 if (ldd != NULL && ldd->lld_sa_generation == lli->lli_sa_generation)
1401 int cl_sync_file_range(struct inode *inode, loff_t start, loff_t end,
1402 enum cl_fsync_mode mode, int ignore_layout);
1404 static inline int ll_file_nolock(const struct file *file)
1406 struct ll_file_data *fd = LUSTRE_FPRIVATE(file);
1407 struct inode *inode = file_inode((struct file *)file);
1409 LASSERT(fd != NULL);
1410 return ((fd->fd_flags & LL_FILE_IGNORE_LOCK) ||
1411 (ll_i2sbi(inode)->ll_flags & LL_SBI_NOLCK));
1414 static inline void ll_set_lock_data(struct obd_export *exp, struct inode *inode,
1415 struct lookup_intent *it, __u64 *bits)
1417 if (!it->it_lock_set) {
1418 struct lustre_handle handle;
1420 /* If this inode is a remote object, it will get two
1421 * separate locks in different namespaces, Master MDT,
1422 * where the name entry is, will grant LOOKUP lock,
1423 * remote MDT, where the object is, will grant
1424 * UPDATE|PERM lock. The inode will be attched to both
1425 * LOOKUP and PERM locks, so revoking either locks will
1426 * case the dcache being cleared */
1427 if (it->it_remote_lock_mode) {
1428 handle.cookie = it->it_remote_lock_handle;
1429 CDEBUG(D_DLMTRACE, "setting l_data to inode "DFID
1430 "(%p) for remote lock %#llx\n",
1431 PFID(ll_inode2fid(inode)), inode,
1433 md_set_lock_data(exp, &handle, inode, NULL);
1436 handle.cookie = it->it_lock_handle;
1438 CDEBUG(D_DLMTRACE, "setting l_data to inode "DFID"(%p)"
1439 " for lock %#llx\n",
1440 PFID(ll_inode2fid(inode)), inode, handle.cookie);
1442 md_set_lock_data(exp, &handle, inode, &it->it_lock_bits);
1443 it->it_lock_set = 1;
1447 *bits = it->it_lock_bits;
1450 static inline void ll_lock_dcache(struct inode *inode)
1452 #ifdef HAVE_DCACHE_LOCK
1453 spin_lock(&dcache_lock);
1455 spin_lock(&inode->i_lock);
1459 static inline void ll_unlock_dcache(struct inode *inode)
1461 #ifdef HAVE_DCACHE_LOCK
1462 spin_unlock(&dcache_lock);
1464 spin_unlock(&inode->i_lock);
1468 static inline int d_lustre_invalid(const struct dentry *dentry)
1470 struct ll_dentry_data *lld = ll_d2d(dentry);
1472 return (lld == NULL) || lld->lld_invalid;
1475 static inline void __d_lustre_invalidate(struct dentry *dentry)
1477 struct ll_dentry_data *lld = ll_d2d(dentry);
1480 lld->lld_invalid = 1;
1484 * Mark dentry INVALID, if dentry refcount is zero (this is normally case for
1485 * ll_md_blocking_ast), unhash this dentry, and let dcache to reclaim it later;
1486 * else dput() of the last refcount will unhash this dentry and kill it.
1488 static inline void d_lustre_invalidate(struct dentry *dentry, int nested)
1490 CDEBUG(D_DENTRY, "invalidate dentry %.*s (%p) parent %p inode %p "
1491 "refc %d\n", dentry->d_name.len, dentry->d_name.name, dentry,
1492 dentry->d_parent, dentry->d_inode, ll_d_count(dentry));
1494 spin_lock_nested(&dentry->d_lock,
1495 nested ? DENTRY_D_LOCK_NESTED : DENTRY_D_LOCK_NORMAL);
1496 __d_lustre_invalidate(dentry);
1498 * We should be careful about dentries created by d_obtain_alias().
1499 * These dentries are not put in the dentry tree, instead they are
1500 * linked to sb->s_anon through dentry->d_hash.
1501 * shrink_dcache_for_umount() shrinks the tree and sb->s_anon list.
1502 * If we unhashed such a dentry, unmount would not be able to find
1503 * it and busy inodes would be reported.
1505 if (ll_d_count(dentry) == 0 && !(dentry->d_flags & DCACHE_DISCONNECTED))
1507 spin_unlock(&dentry->d_lock);
1510 static inline void d_lustre_revalidate(struct dentry *dentry)
1512 spin_lock(&dentry->d_lock);
1513 LASSERT(ll_d2d(dentry) != NULL);
1514 ll_d2d(dentry)->lld_invalid = 0;
1515 spin_unlock(&dentry->d_lock);
1518 static inline dev_t ll_compat_encode_dev(dev_t dev)
1520 /* The compat_sys_*stat*() syscalls will fail unless the
1521 * device majors and minors are both less than 256. Note that
1522 * the value returned here will be passed through
1523 * old_encode_dev() in cp_compat_stat(). And so we are not
1524 * trying to return a valid compat (u16) device number, just
1525 * one that will pass the old_valid_dev() check. */
1527 return MKDEV(MAJOR(dev) & 0xff, MINOR(dev) & 0xff);
1530 int ll_layout_conf(struct inode *inode, const struct cl_object_conf *conf);
1531 int ll_layout_refresh(struct inode *inode, __u32 *gen);
1532 int ll_layout_restore(struct inode *inode, loff_t start, __u64 length);
1533 int ll_layout_write_intent(struct inode *inode, enum layout_intent_opc opc,
1534 struct lu_extent *ext);
1536 int ll_xattr_init(void);
1537 void ll_xattr_fini(void);
1539 int ll_page_sync_io(const struct lu_env *env, struct cl_io *io,
1540 struct cl_page *page, enum cl_req_type crt);
1542 int ll_getparent(struct file *file, struct getparent __user *arg);
1545 int cl_setattr_ost(struct cl_object *obj, const struct iattr *attr,
1546 enum op_xvalid xvalid, unsigned int attr_flags);
1548 extern struct lu_env *cl_inode_fini_env;
1549 extern __u16 cl_inode_fini_refcheck;
1551 int cl_file_inode_init(struct inode *inode, struct lustre_md *md);
1552 void cl_inode_fini(struct inode *inode);
1554 u64 cl_fid_build_ino(const struct lu_fid *fid, int api32);
1555 u32 cl_fid_build_gen(const struct lu_fid *fid);
1557 #endif /* LLITE_INTERNAL_H */