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/
32 #ifndef LLITE_INTERNAL_H
33 #define LLITE_INTERNAL_H
35 #include <lustre_disk.h> /* for s2sbi */
36 #include <lustre_linkea.h>
38 /* for struct cl_lock_descr and struct cl_io */
39 #include <cl_object.h>
40 #include <lustre_lmv.h>
41 #include <lustre_mdc.h>
42 #include <lustre_intent.h>
43 #include <linux/compat.h>
44 #include <linux/aio.h>
45 #include <linux/parser.h>
46 #include <linux/seqlock.h>
47 #include <lustre_compat.h>
48 #include <lustre_crypto.h>
49 #include <range_lock.h>
51 #include "vvp_internal.h"
53 #include "foreign_symlink.h"
59 #ifndef HAVE_VM_FAULT_RETRY
60 #define VM_FAULT_RETRY 0
63 /* Kernel 3.1 kills LOOKUP_CONTINUE, LOOKUP_PARENT is equivalent to it.
64 * seem kernel commit 49084c3bb2055c401f3493c13edae14d49128ca0
66 #ifndef LOOKUP_CONTINUE
67 #define LOOKUP_CONTINUE LOOKUP_PARENT
70 #ifdef HAVE_DENTRY_D_U_D_ALIAS
71 #define d_alias d_u.d_alias
74 /** Only used on client-side for indicating the tail of dir hash/offset. */
75 #define LL_DIR_END_OFF 0x7fffffffffffffffULL
76 #define LL_DIR_END_OFF_32BIT 0x7fffffffUL
78 /* 4UL * 1024 * 1024 */
79 #define LL_MAX_BLKSIZE_BITS 22
81 #define LL_IT2STR(it) ((it) ? ldlm_it2str((it)->it_op) : "0")
83 #define TIMES_SET_FLAGS (ATTR_MTIME_SET | ATTR_ATIME_SET | ATTR_TIMES_SET)
85 struct ll_dentry_data {
86 unsigned int lld_sa_generation;
87 unsigned int lld_invalid:1;
88 struct rcu_head lld_rcu_head;
91 #define ll_d2d(de) ((struct ll_dentry_data *)((de)->d_fsdata))
93 #define LLI_INODE_MAGIC 0x111d0de5
94 #define LLI_INODE_DEAD 0xdeadd00d
96 struct ll_getname_data {
97 #ifdef HAVE_DIR_CONTEXT
98 struct dir_context ctx;
100 char *lgd_name; /* points to buf with NAME_MAX+1 size */
101 struct lu_fid lgd_fid; /* target fid we are looking for */
102 int lgd_found; /* inode matched? */
105 struct ll_grouplock {
106 struct lu_env *lg_env;
108 struct cl_lock *lg_lock;
109 unsigned long lg_gid;
112 /* See comment on trunc_sem_down_read_nowait */
113 struct ll_trunc_sem {
114 /* when positive, this is a count of readers, when -1, it indicates
115 * the semaphore is held for write, and 0 is unlocked
117 atomic_t ll_trunc_readers;
118 /* this tracks a count of waiting writers */
119 atomic_t ll_trunc_waiters;
123 LSA_PATTERN_NONE = 0x0000,
124 /* once detected, and found no statahead pattern matched */
125 LSA_PATTERN_INVALID = 0x0001,
126 /* do the directory listing. i.e ls $dir */
127 LSA_PATTERN_LIST = 0x0002,
128 /* regularized file name scanning, i.e. mdtest.$i */
129 LSA_PATTERN_FNAME = 0x0004,
130 /* statahead advise via statahead hint from users */
131 LSA_PATTERN_ADVISE = 0x0008,
132 /* have called opendir() on the parent dir */
133 LSA_PATTERN_OPENDIR = 0x0010,
134 /* not first dirent, or is "." for listing */
135 LSA_PATTERN_LS_NOT_FIRST_DE = 0x0100,
136 /* the file names of stat() calls has regularized predictable format */
137 LSA_PATTERN_FN_PREDICT = 0x1000,
138 /* fname statahead workload similar to mdtest shared dir stat() */
139 LSA_PATTERN_FN_SHARED = 0x2000,
140 /* fname statahead workload similar to mdtest unique dir stat() */
141 LSA_PATTERN_FN_UNIQUE = 0x4000,
142 /* fname statahead workload with stride regularized naming format */
143 LSA_PATTERN_FN_STRIDE = 0x8000,
144 LSA_PATTERN_MASK = (LSA_PATTERN_LIST |
150 struct ll_inode_info {
151 __u32 lli_inode_magic;
154 volatile unsigned long lli_flags;
155 struct posix_acl *lli_posix_acl;
157 /* identifying fields for both metadata and data stacks. */
158 struct lu_fid lli_fid;
159 /* master inode fid for stripe directory */
160 struct lu_fid lli_pfid;
162 /* Need all three because every inode may be opened in different mode */
163 struct obd_client_handle *lli_mds_read_och;
164 struct obd_client_handle *lli_mds_write_och;
165 struct obd_client_handle *lli_mds_exec_och;
166 __u64 lli_open_fd_read_count;
167 __u64 lli_open_fd_write_count;
168 __u64 lli_open_fd_exec_count;
170 /* Protects access to och pointers and their usage counters */
171 struct mutex lli_och_mutex;
173 struct inode lli_vfs_inode;
175 /* the most recent timestamps obtained from mds */
180 spinlock_t lli_agl_lock;
181 bool lli_synced_to_mds;
183 /* inode specific open lock caching threshold */
184 u32 lli_open_thrsh_count;
185 /* Number of times this inode was opened */
186 u64 lli_open_fd_count;
187 /* When last close was performed on this inode */
188 ktime_t lli_close_fd_time;
190 /* Try to make the d::member and f::member are aligned. Before using
191 * these members, make clear whether it is directory or not.
196 /* metadata statahead */
197 /* since parent-child threads can share the same @file
198 * struct, "opendir_key" is the token when dir close for
199 * case of parent exit before child -- it is me should
200 * cleanup the dir readahead.
202 void *lli_opendir_key;
203 struct ll_statahead_info *lli_sai;
204 struct ll_statahead_context *lli_sax;
205 /* protect statahead stuff. */
206 spinlock_t lli_sa_lock;
207 /* "stat_pid" is the token when lookup/revalid
208 * -- I am the owner of dir statahead.
211 /* directory depth to ROOT */
212 unsigned short lli_dir_depth;
213 /* directory depth to ancestor whose default LMV is
216 unsigned short lli_inherit_depth;
217 /* stat will try to access statahead entries or start
218 * statahead if this flag is set, and this flag will be
219 * set upon dir open, and cleared when dir is closed,
220 * statahead hit ratio is too low, or start statahead
223 unsigned short lli_sa_enabled:1,
224 /* default LMV is explicitly set in inode on MDT, this
225 * is for old server, or default LMV is set by
226 * "lfs setdirstripe -D".
228 lli_def_lsm_obj_set:1;
229 /* generation for statahead */
230 unsigned int lli_sa_generation;
231 /* access pattern for statahead */
232 enum ll_sa_pattern lli_sa_pattern;
234 * suffix index number of the latest stat dentry. It
235 * is used for the detection of the file name statahead
238 __u32 lli_sa_fname_index;
240 * indicate the count that the suffix index number
241 * matched continuously. This field is using for the
242 * detection of the file name statahead pattern.
244 unsigned int lli_sa_match_count;
245 /* rw lock protects lli_lsm_md */
246 struct rw_semaphore lli_lsm_sem;
247 /* directory stripe information */
248 struct lmv_stripe_object *lli_lsm_obj;
249 /* directory default LMV */
250 struct lmv_stripe_object *lli_def_lsm_obj;
253 /* for non-directory */
255 struct mutex lli_size_mutex;
256 struct task_struct *lli_size_lock_owner;
257 char *lli_symlink_name;
258 struct ll_trunc_sem lli_trunc_sem;
259 struct range_lock_tree lli_write_tree;
260 struct mutex lli_setattr_mutex;
262 struct rw_semaphore lli_glimpse_sem;
263 ktime_t lli_glimpse_time;
264 struct list_head lli_agl_list;
267 /* for writepage() only to communicate to fsync */
270 /* protect the file heat fields */
271 spinlock_t lli_heat_lock;
272 __u32 lli_heat_flags;
273 struct obd_heat_instance lli_heat_instances[OBD_HEAT_COUNT];
276 * Whenever a process try to read/write the file, the
277 * jobid, uid and gid of the process will be saved here,
278 * and it'll be packed into write RPC when flush later.
280 * So the read/write statistics or TBF rules for jobid,
281 * uid or gid will not be accurate if the file is shared
284 char lli_jobid[LUSTRE_JOBID_SIZE];
288 struct mutex lli_pcc_lock;
289 enum lu_pcc_state_flags lli_pcc_state;
291 * @lli_pcc_generation saves the gobal PCC generation
292 * when the file was successfully attached into PCC.
293 * The flags of the PCC dataset are saved in
295 * The gobal PCC generation will be increased when add
296 * or delete a PCC backend, or change the configuration
297 * parameters for PCC.
298 * If @lli_pcc_generation is same as the gobal PCC
299 * generation, we can use the saved flags of the PCC
300 * dataset to determine whether need to try auto attach
303 __u64 lli_pcc_generation;
304 enum pcc_dataset_flags lli_pcc_dsflags;
305 struct pcc_inode *lli_pcc_inode;
307 struct mutex lli_group_mutex;
308 __u64 lli_group_users;
309 unsigned long lli_group_gid;
311 __u64 lli_attr_valid;
313 __u64 lli_lazyblocks;
317 /* XXX: For following frequent used members, although they maybe special
318 * used for non-directory object, it is some time-wasting to check
319 * whether the object is directory or not before using them. On the
320 * other hand, currently, sizeof(f) > sizeof(d), it cannot reduce
321 * the "ll_inode_info" size even if moving those members into u.f.
322 * So keep them out side.
324 * In the future, if more members are added only for directory,
325 * some of the following members can be moved into u.f.
327 struct cl_object *lli_clob;
329 /* mutex to request for layout lock exclusively. */
330 struct mutex lli_layout_mutex;
331 struct task_struct *lli_layout_lock_owner;
332 /* Layout version, protected by lli_layout_lock */
333 __u32 lli_layout_gen;
334 spinlock_t lli_layout_lock;
336 __u32 lli_projid; /* project id */
338 struct rw_semaphore lli_xattrs_list_rwsem;
339 struct mutex lli_xattrs_enq_lock;
340 struct list_head lli_xattrs; /* ll_xattr_entry->xe_list*/
341 struct list_head lli_lccs; /* list of ll_cl_context */
342 seqlock_t lli_page_inv_lock;
344 struct task_struct *lli_inode_lock_owner;
347 #ifndef HAVE_USER_NAMESPACE_ARG
348 #define inode_permission(ns, inode, mask) inode_permission(inode, mask)
349 #define generic_permission(ns, inode, mask) generic_permission(inode, mask)
350 #define simple_setattr(ns, de, iattr) simple_setattr(de, iattr)
351 #define ll_setattr(ns, de, attr) ll_setattr(de, attr)
352 #ifdef HAVE_STRUCT_POSIX_ACL_XATTR
353 #define setattr_prepare(ns, de, at) setattr_prepare(de, at)
355 #define setattr_prepare(ns, de, at) inode_change_ok(de->d_inode, at)
357 #define ll_inode_permission(ns, inode, mask) ll_inode_permission(inode, mask)
358 #ifdef HAVE_INODEOPS_ENHANCED_GETATTR
359 #define ll_getattr(ns, path, stat, mask, fl) ll_getattr(path, stat, mask, fl)
360 #endif /* HAVE_INODEOPS_ENHANCED_GETATTR */
364 #define iocb_ki_flags_check(flag, name) (!!((flag) & IOCB_ ## name))
365 #define ki_flag(name) IOCB_ ## name
367 #define iocb_ki_flags_check(flag, name) (!!((flag) & O_ ## name))
368 #define ki_flag(name) O_ ## name
371 static inline void ll_trunc_sem_init(struct ll_trunc_sem *sem)
373 atomic_set(&sem->ll_trunc_readers, 0);
374 atomic_set(&sem->ll_trunc_waiters, 0);
377 /* This version of down read ignores waiting writers, meaning if the semaphore
378 * is already held for read, this down_read will 'join' that reader and also
379 * take the semaphore.
381 * This lets us avoid an unusual deadlock.
383 * We must take lli_trunc_sem in read mode on entry in to various i/o paths
384 * in Lustre, in order to exclude truncates. Some of these paths then need to
385 * take the mmap_lock, while still holding the trunc_sem. The problem is that
386 * page faults hold the mmap_lock when calling in to Lustre, and then must also
387 * take the trunc_sem to exclude truncate.
389 * This means the locking order for trunc_sem and mmap_lock is sometimes AB,
390 * sometimes BA. This is almost OK because in both cases, we take the trunc
391 * sem for read, so it doesn't block.
393 * However, if a write mode user (truncate, a setattr op) arrives in the
394 * middle of this, the second reader on the truncate_sem will wait behind that
397 * So we have, on our truncate sem, in order (where 'reader' and 'writer' refer
398 * to the mode in which they take the semaphore):
399 * reader (holding mmap_lock, needs truncate_sem)
401 * reader (holding truncate sem, waiting for mmap_lock)
403 * And so the readers deadlock.
405 * The solution is this modified semaphore, where this down_read ignores
406 * waiting write operations, and all waiters are woken up at once, so readers
407 * using down_read_nowait cannot get stuck behind waiting writers, regardless
408 * of the order they arrived in.
410 * down_read_nowait is only used in the page fault case, where we already hold
411 * the mmap_lock. This is because otherwise repeated read and write operations
412 * (which take the truncate sem) could prevent a truncate from ever starting.
413 * This could still happen with page faults, but without an even more complex
414 * mechanism, this is unavoidable.
418 static inline void trunc_sem_down_read_nowait(struct ll_trunc_sem *sem)
420 wait_var_event(&sem->ll_trunc_readers,
421 atomic_inc_unless_negative(&sem->ll_trunc_readers));
424 static inline void trunc_sem_down_read(struct ll_trunc_sem *sem)
426 wait_var_event(&sem->ll_trunc_readers,
427 atomic_read(&sem->ll_trunc_waiters) == 0 &&
428 atomic_inc_unless_negative(&sem->ll_trunc_readers));
431 static inline void trunc_sem_up_read(struct ll_trunc_sem *sem)
433 if (atomic_dec_return(&sem->ll_trunc_readers) == 0 &&
434 atomic_read(&sem->ll_trunc_waiters))
435 wake_up_var(&sem->ll_trunc_readers);
438 static inline void trunc_sem_down_write(struct ll_trunc_sem *sem)
440 atomic_inc(&sem->ll_trunc_waiters);
441 wait_var_event(&sem->ll_trunc_readers,
442 atomic_cmpxchg(&sem->ll_trunc_readers, 0, -1) == 0);
443 atomic_dec(&sem->ll_trunc_waiters);
446 static inline void trunc_sem_up_write(struct ll_trunc_sem *sem)
448 atomic_set(&sem->ll_trunc_readers, 0);
449 /* match the smp_mb() in wait_var_event()->prepare_to_wait() */
451 wake_up_var(&sem->ll_trunc_readers);
454 #ifdef CONFIG_LUSTRE_FS_POSIX_ACL
455 static inline void lli_clear_acl(struct ll_inode_info *lli)
457 if (lli->lli_posix_acl) {
458 posix_acl_release(lli->lli_posix_acl);
459 lli->lli_posix_acl = NULL;
463 static inline void lli_replace_acl(struct ll_inode_info *lli,
464 struct lustre_md *md)
466 write_lock(&lli->lli_lock);
467 if (lli->lli_posix_acl)
468 posix_acl_release(lli->lli_posix_acl);
469 lli->lli_posix_acl = md->posix_acl;
470 write_unlock(&lli->lli_lock);
473 static inline void lli_clear_acl(struct ll_inode_info *lli)
477 static inline void lli_replace_acl(struct ll_inode_info *lli,
478 struct lustre_md *md)
483 static inline __u32 ll_layout_version_get(struct ll_inode_info *lli)
487 spin_lock(&lli->lli_layout_lock);
488 gen = lli->lli_layout_gen;
489 spin_unlock(&lli->lli_layout_lock);
494 static inline void ll_layout_version_set(struct ll_inode_info *lli, __u32 gen)
496 spin_lock(&lli->lli_layout_lock);
497 lli->lli_layout_gen = gen;
498 spin_unlock(&lli->lli_layout_lock);
501 enum ll_inode_flags {
502 /* File data is modified. */
503 LLIF_DATA_MODIFIED = 0,
504 /* File is being restored */
505 LLIF_FILE_RESTORING = 1,
506 /* Xattr cache is attached to the file */
507 LLIF_XATTR_CACHE = 2,
508 /* Project inherit */
509 LLIF_PROJECT_INHERIT = 3,
510 /* update atime from MDS even if it's older than local inode atime. */
511 LLIF_UPDATE_ATIME = 4,
512 /* foreign file/dir can be unlinked unconditionnaly */
513 LLIF_FOREIGN_REMOVABLE = 5,
514 /* 6 is not used for now */
515 /* Xattr cache is filled */
516 LLIF_XATTR_CACHE_FILLED = 7,
520 int ll_xattr_cache_destroy(struct inode *inode);
521 int ll_xattr_cache_empty(struct inode *inode);
523 int ll_xattr_cache_get(struct inode *inode,
529 int ll_xattr_cache_insert(struct inode *inode,
534 static inline bool obd_connect_has_secctx(struct obd_connect_data *data)
536 #ifdef CONFIG_SECURITY
537 return data->ocd_connect_flags & OBD_CONNECT_FLAGS2 &&
538 data->ocd_connect_flags2 & OBD_CONNECT2_FILE_SECCTX;
544 static inline void obd_connect_set_secctx(struct obd_connect_data *data)
546 #ifdef CONFIG_SECURITY
547 data->ocd_connect_flags2 |= OBD_CONNECT2_FILE_SECCTX;
551 /* Only smack and selinux is known to use security contexts */
552 static inline bool ll_xattr_is_seclabel(const char *name)
554 return !strcmp(name, XATTR_NAME_SELINUX) ||
555 !strcmp(name, XATTR_NAME_SMACK);
558 static inline bool ll_xattr_suffix_is_seclabel(const char *suffix)
560 return !strcmp(suffix, XATTR_SELINUX_SUFFIX) ||
561 !strcmp(suffix, XATTR_SMACK_SUFFIX);
564 int ll_dentry_init_security(struct dentry *dentry, int mode, struct qstr *name,
565 const char **secctx_name, __u32 *secctx_name_size,
566 void **secctx, __u32 *secctx_size,
569 int ll_inode_init_security(struct dentry *dentry, struct inode *inode,
572 int ll_inode_notifysecctx(struct inode *inode,
573 void *secctx, __u32 secctxlen);
575 void ll_secctx_name_free(struct ll_sb_info *sbi);
577 int ll_secctx_name_store(struct inode *in);
579 __u32 ll_secctx_name_get(struct ll_sb_info *sbi, const char **secctx_name);
581 int ll_security_secctx_name_filter(struct ll_sb_info *sbi, int xattr_type,
584 static inline bool obd_connect_has_unaligned_dio(struct obd_connect_data *data)
586 return data->ocd_connect_flags & OBD_CONNECT_FLAGS2 &&
587 data->ocd_connect_flags2 & OBD_CONNECT2_UNALIGNED_DIO;
590 static inline bool obd_connect_has_enc(struct obd_connect_data *data)
592 #ifdef HAVE_LUSTRE_CRYPTO
593 return data->ocd_connect_flags & OBD_CONNECT_FLAGS2 &&
594 data->ocd_connect_flags2 & OBD_CONNECT2_ENCRYPT;
600 static inline void obd_connect_set_enc(struct obd_connect_data *data)
602 #ifdef HAVE_LUSTRE_CRYPTO
603 data->ocd_connect_flags2 |= OBD_CONNECT2_ENCRYPT;
607 static inline bool obd_connect_has_name_enc(struct obd_connect_data *data)
609 #ifdef HAVE_LUSTRE_CRYPTO
610 return data->ocd_connect_flags & OBD_CONNECT_FLAGS2 &&
611 data->ocd_connect_flags2 & OBD_CONNECT2_ENCRYPT_NAME;
617 static inline void obd_connect_set_name_enc(struct obd_connect_data *data)
619 #ifdef HAVE_LUSTRE_CRYPTO
620 data->ocd_connect_flags2 |= OBD_CONNECT2_ENCRYPT_NAME;
624 static inline bool obd_connect_has_enc_fid2path(struct obd_connect_data *data)
626 #ifdef HAVE_LUSTRE_CRYPTO
627 return data->ocd_connect_flags & OBD_CONNECT_FLAGS2 &&
628 data->ocd_connect_flags2 & OBD_CONNECT2_ENCRYPT_FID2PATH;
634 static inline void obd_connect_set_enc_fid2path(struct obd_connect_data *data)
636 #ifdef HAVE_LUSTRE_CRYPTO
637 data->ocd_connect_flags2 |= OBD_CONNECT2_ENCRYPT_FID2PATH;
642 * Locking to guarantee consistency of non-atomic updates to long long i_size,
643 * consistency between file size and KMS.
645 * Implemented by ->lli_size_mutex and ->lsm_lock, nested in that order.
647 void ll_inode_size_lock(struct inode *inode);
648 void ll_inode_size_unlock(struct inode *inode);
649 int ll_inode_size_trylock(struct inode *inode);
651 static inline struct ll_inode_info *ll_i2info(struct inode *inode)
653 return container_of(inode, struct ll_inode_info, lli_vfs_inode);
656 static inline struct pcc_inode *ll_i2pcci(struct inode *inode)
658 return ll_i2info(inode)->lli_pcc_inode;
661 static inline void ll_set_inode_lock_owner(struct inode *inode)
663 ll_i2info(inode)->lli_inode_lock_owner = current;
666 static inline void ll_clear_inode_lock_owner(struct inode *inode)
668 ll_i2info(inode)->lli_inode_lock_owner = NULL;
671 static inline struct task_struct *ll_get_inode_lock_owner(struct inode *inode)
673 return ll_i2info(inode)->lli_inode_lock_owner;
676 /* lock inode and set inode lock owener */
677 static inline void ll_inode_lock(struct inode *inode)
680 ll_set_inode_lock_owner(inode);
683 /* clear inode lock owner and unlock it */
684 static inline void ll_inode_unlock(struct inode *inode)
686 ll_clear_inode_lock_owner(inode);
690 /* default to use at least 16M for fast read if possible */
691 #define RA_REMAIN_WINDOW_MIN MiB_TO_PAGES(16UL)
693 /* default read-ahead on a given client mountpoint. */
694 #define SBI_DEFAULT_READ_AHEAD_MAX MiB_TO_PAGES(1024UL)
695 /* on small nodes (ie, testing VMs), we need at least this much to make
696 * readahead easily testable
698 #define SBI_DEFAULT_READ_AHEAD_MIN MiB_TO_PAGES(256UL)
700 /* default read-ahead for a single file descriptor */
701 #define SBI_DEFAULT_READ_AHEAD_PER_FILE_MAX MiB_TO_PAGES(256UL)
703 /* default read-ahead full files smaller than limit on the second read */
704 #define SBI_DEFAULT_READ_AHEAD_WHOLE_MAX MiB_TO_PAGES(2UL)
706 /* default range pages */
707 #define SBI_DEFAULT_RA_RANGE_PAGES MiB_TO_PAGES(1ULL)
709 /* Min range pages */
710 #define RA_MIN_MMAP_RANGE_PAGES 16UL
715 RA_STAT_DISTANT_READPAGE,
716 RA_STAT_MISS_IN_WINDOW,
717 RA_STAT_FAILED_GRAB_PAGE,
718 RA_STAT_FAILED_MATCH,
723 RA_STAT_MAX_IN_FLIGHT,
724 RA_STAT_WRONG_GRAB_PAGE,
725 RA_STAT_FAILED_REACH_END,
727 RA_STAT_FAILED_FAST_READ,
728 RA_STAT_MMAP_RANGE_READ,
729 RA_STAT_READAHEAD_PAGES,
734 atomic_t ra_cur_pages;
735 unsigned long ra_max_pages;
736 unsigned long ra_max_pages_per_file;
737 unsigned long ra_range_pages;
738 unsigned long ra_max_read_ahead_whole_pages;
739 struct workqueue_struct *ll_readahead_wq;
741 * Max number of active works could be triggered
742 * for async readahead.
744 unsigned int ra_async_max_active;
745 /* how many async readahead triggered in flight */
746 atomic_t ra_async_inflight;
747 /* Threshold to control when to trigger async readahead */
748 unsigned long ra_async_pages_per_file_threshold;
751 /* ra_io_arg will be filled in the beginning of ll_readahead with
752 * ras_lock, then the following ll_read_ahead_pages will read RA
753 * pages according to this arg, all the items in this structure are
754 * counted by page index.
757 pgoff_t ria_start_idx; /* start offset of read-ahead*/
758 pgoff_t ria_end_idx; /* end offset of read-ahead*/
759 unsigned long ria_reserved; /* reserved pages for read-ahead */
760 pgoff_t ria_end_idx_min;/* minimum end to cover current read */
761 bool ria_eof; /* reach end of file */
762 /* If stride read pattern is detected, ria_stoff is the byte offset
763 * where stride read is started. Note: for normal read-ahead, the
764 * value here is meaningless, and also it will not be accessed
767 /* ria_length and ria_bytes are the length and pages length in the
768 * stride I/O mode. And they will also be used to check whether
769 * it is stride I/O read-ahead in the read-ahead pages
775 /* LL_HIST_MAX=32 causes an overflow */
776 #define LL_HIST_MAX 28
777 #define LL_HIST_START 12 /* buckets start at 2^12 = 4k */
778 #define LL_PROCESS_HIST_MAX 10
779 struct per_process_info {
781 struct obd_histogram pp_r_hist;
782 struct obd_histogram pp_w_hist;
785 /* pp_extents[LL_PROCESS_HIST_MAX] will hold the combined process info */
786 struct ll_rw_extents_info {
788 struct per_process_info pp_extents[LL_PROCESS_HIST_MAX + 1];
791 #define LL_OFFSET_HIST_MAX 100
792 struct ll_rw_process_info {
795 loff_t rw_range_start;
797 loff_t rw_last_file_pos;
799 size_t rw_smallest_extent;
800 size_t rw_largest_extent;
801 struct ll_file_data *rw_last_file;
804 enum stats_track_type {
805 STATS_TRACK_ALL = 0, /* track all processes */
806 STATS_TRACK_PID, /* track process with this pid */
807 STATS_TRACK_PPID, /* track processes with this ppid */
808 STATS_TRACK_GID, /* track processes with this gid */
812 /* flags for sbi->ll_flags */
814 LL_SBI_NOLCK, /* DLM locking disabled directio-only */
815 LL_SBI_CHECKSUM, /* checksum each page as it's written */
816 LL_SBI_LOCALFLOCK, /* local flocks instead of fs-wide */
817 LL_SBI_FLOCK, /* flock enabled */
818 LL_SBI_USER_XATTR, /* support user xattr */
819 LL_SBI_LRU_RESIZE, /* lru resize support */
820 LL_SBI_LAZYSTATFS, /* lazystatfs mount option */
821 LL_SBI_32BIT_API, /* generate 32 bit inodes. */
822 LL_SBI_USER_FID2PATH, /* fid2path by unprivileged users */
823 LL_SBI_VERBOSE, /* verbose mount/umount */
824 LL_SBI_ALWAYS_PING, /* ping even if server suppress_pings */
825 LL_SBI_TEST_DUMMY_ENCRYPTION, /* test dummy encryption */
826 LL_SBI_ENCRYPT, /* client side encryption */
827 LL_SBI_FOREIGN_SYMLINK, /* foreign fake-symlink support */
828 LL_SBI_FOREIGN_SYMLINK_UPCALL, /* foreign fake-symlink upcall set */
829 LL_SBI_STATFS_PROJECT, /* statfs returns project quota */
830 LL_SBI_NUM_MOUNT_OPT,
832 LL_SBI_ACL, /* support ACL */
833 LL_SBI_AGL_ENABLED, /* enable agl */
834 LL_SBI_64BIT_HASH, /* support 64-bits dir hash/offset */
835 LL_SBI_LAYOUT_LOCK, /* layout lock support */
836 LL_SBI_XATTR_CACHE, /* support for xattr cache */
837 LL_SBI_NOROOTSQUASH, /* do not apply root squash */
838 LL_SBI_FAST_READ, /* fast read support */
839 LL_SBI_FILE_SECCTX, /* file security context at create */
840 LL_SBI_TINY_WRITE, /* tiny write support */
841 LL_SBI_FILE_HEAT, /* file heat support */
842 LL_SBI_PARALLEL_DIO, /* parallel (async) O_DIRECT RPCs */
843 LL_SBI_ENCRYPT_NAME, /* name encryption */
844 LL_SBI_UNALIGNED_DIO, /* unaligned DIO */
845 LL_SBI_HYBRID_IO, /* allow BIO as DIO */
849 int ll_sbi_flags_seq_show(struct seq_file *m, void *v);
851 /* This is embedded into llite super-blocks to keep track of connect
852 * flags (capabilities) supported by all imports given mount is
855 struct lustre_client_ocd {
856 /* This is conjunction of connect_flags across all imports
857 * (LOVs) this mount is connected to. This field is updated by
858 * cl_ocd_update() under ->lco_lock.
861 struct mutex lco_lock;
862 struct obd_export *lco_md_exp;
863 struct obd_export *lco_dt_exp;
867 /* this protects pglist and ra_info. It isn't safe to
868 * grab from interrupt contexts
871 spinlock_t ll_pp_extent_lock; /* pp_extent entry*/
872 spinlock_t ll_process_lock; /* ll_rw_process_info */
873 struct lustre_sb_info *lsi;
874 struct obd_uuid ll_sb_uuid;
875 struct obd_export *ll_md_exp;
876 struct obd_export *ll_dt_exp;
877 struct obd_device *ll_md_obd;
878 struct obd_device *ll_dt_obd;
879 struct dentry *ll_debugfs_entry;
880 struct lu_fid ll_root_fid; /* root object fid */
881 struct mnt_namespace *ll_mnt_ns;
883 DECLARE_BITMAP(ll_flags, LL_SBI_NUM_FLAGS); /* enum ll_sbi_flags */
884 unsigned int ll_xattr_cache_enabled:1,
885 ll_xattr_cache_set:1, /* already set to 0/1 */
886 ll_client_common_fill_super_succeeded:1,
888 ll_inode_cache_enabled:1,
889 ll_enable_statahead_fname:1,
890 ll_intent_mkdir_enabled:1;
892 struct lustre_client_ocd ll_lco;
894 struct lprocfs_stats *ll_stats; /* lprocfs stats counter */
896 /* Used to track "unstable" pages on a client, and maintain a
897 * LRU list of clean pages. An "unstable" page is defined as
898 * any page which is sent to a server as part of a bulk request,
899 * but is uncommitted to stable storage.
901 struct cl_client_cache *ll_cache;
903 struct lprocfs_stats *ll_ra_stats;
905 struct ll_ra_info ll_ra_info;
906 unsigned int ll_namelen;
907 const struct file_operations *ll_fop;
909 struct lu_site *ll_site;
910 struct cl_device *ll_cl;
913 struct ll_rw_extents_info *ll_rw_extents_info;
914 int ll_extent_process_count;
915 unsigned int ll_offset_process_count;
916 struct ll_rw_process_info *ll_rw_process_info;
917 struct ll_rw_process_info *ll_rw_offset_info;
918 ktime_t ll_process_stats_init;
919 unsigned int ll_rw_offset_entry_count;
920 int ll_stats_track_id;
921 enum stats_track_type ll_stats_track_type;
924 /* metadata stat-ahead */
925 unsigned int ll_sa_running_max;/* max concurrent sa instant */
926 /* max SUB request cont in a batch PTLRPC request */
927 unsigned int ll_sa_batch_max;
928 unsigned int ll_sa_max; /* max statahead RPCs */
929 unsigned int ll_sa_min; /* min statahead req count */
930 atomic_t ll_sa_total; /* sa thread started count */
931 atomic_t ll_sa_wrong; /* sa stopped low hit ratio */
932 atomic_t ll_sa_running; /* running sa thread count */
933 atomic_t ll_agl_total; /* AGL thread started count */
934 atomic_t ll_sa_hit_total; /* total hit count */
935 atomic_t ll_sa_miss_total; /* total miss count */
936 /* statahead thread count started for directory traversing pattern. */
937 atomic_t ll_sa_list_total;
938 /* statahead thread count started for regularized file name pattern. */
939 atomic_t ll_sa_fname_total;
941 * stop the statahead thread if it is not doing a stat() in such time
942 * period as it probably does not care too much about performance or
943 * the user is no longer using this directory.
945 unsigned long ll_sa_timeout;
946 /* save s_dev before assign for clustred nfs */
949 struct root_squash_info ll_squash;
952 /* st_blksize returned by stat(2), when non-zero */
953 unsigned int ll_stat_blksize;
955 /* maximum relative age of cached statfs results */
956 unsigned int ll_statfs_max_age;
958 struct kset ll_kset; /* sysfs object */
959 struct completion ll_kobj_unregister;
962 unsigned int ll_heat_decay_weight;
963 unsigned int ll_heat_period_second;
965 /* Opens of the same inode before we start requesting open lock */
966 u32 ll_oc_thrsh_count;
968 /* Time in ms between last inode close and next open to be considered
969 * instant back to back and would trigger an open lock request
973 /* Time in ms after last file close we no longer count prior opens*/
976 /* filesystem fsname */
977 char ll_fsname[LUSTRE_MAXFSNAME + 1];
979 /* Persistent Client Cache */
980 struct pcc_super ll_pcc_super;
982 /* to protect vs updates in all following foreign symlink fields */
983 struct rw_semaphore ll_foreign_symlink_sem;
984 /* foreign symlink path prefix */
985 char *ll_foreign_symlink_prefix;
986 /* full prefix size including leading '\0' */
987 size_t ll_foreign_symlink_prefix_size;
988 /* foreign symlink path upcall */
989 char *ll_foreign_symlink_upcall;
990 /* foreign symlink path upcall infos */
991 struct ll_foreign_symlink_upcall_item *ll_foreign_symlink_upcall_items;
992 /* foreign symlink path upcall nb infos */
993 unsigned int ll_foreign_symlink_upcall_nb_items;
995 /* cached file security context xattr name. e.g: security.selinux */
996 char *ll_secctx_name;
997 __u32 ll_secctx_name_size;
999 /* LU-14535: the list of "lfs quota -a" */
1000 struct list_head ll_all_quota_list;
1003 #define SBI_DEFAULT_HEAT_DECAY_WEIGHT ((80 * 256 + 50) / 100)
1004 #define SBI_DEFAULT_HEAT_PERIOD_SECOND (60)
1006 #define SBI_DEFAULT_OPENCACHE_THRESHOLD_COUNT (5)
1007 #define SBI_DEFAULT_OPENCACHE_THRESHOLD_MS (100) /* 0.1 second */
1008 #define SBI_DEFAULT_OPENCACHE_THRESHOLD_MAX_MS (60000) /* 1 minute */
1010 /* per file-descriptor read-ahead data. */
1011 struct ll_readahead_state {
1012 spinlock_t ras_lock;
1013 /* End byte that read(2) try to read. */
1014 loff_t ras_last_read_end_bytes;
1016 * number of bytes read after last read-ahead window reset. As window
1017 * is reset on each seek, this is effectively a number of consecutive
1018 * accesses. Maybe ->ras_accessed_in_window is better name.
1020 * XXX nikita: window is also reset (by ras_update()) when Lustre
1021 * believes that memory pressure evicts read-ahead pages. In that
1022 * case, it probably doesn't make sense to expand window to
1023 * PTLRPC_MAX_BRW_PAGES on the third access.
1025 loff_t ras_consecutive_bytes;
1027 * number of read requests after the last read-ahead window reset
1028 * As window is reset on each seek, this is effectively the number
1029 * on consecutive read request and is used to trigger read-ahead.
1031 unsigned long ras_consecutive_requests;
1033 * Parameters of current read-ahead window. Handled by
1034 * ras_update(). On the initial access to the file or after a seek,
1035 * window is reset to 0. After 3 consecutive accesses, window is
1036 * expanded to PTLRPC_MAX_BRW_PAGES. Afterwards, window is enlarged by
1037 * PTLRPC_MAX_BRW_PAGES chunks up to ->ra_max_pages.
1039 pgoff_t ras_window_start_idx;
1040 pgoff_t ras_window_pages;
1042 /* Page index where min range read starts */
1043 pgoff_t ras_range_min_start_idx;
1044 /* Page index where mmap range read ends */
1045 pgoff_t ras_range_max_end_idx;
1046 /* number of mmap pages where last time detected */
1047 pgoff_t ras_last_range_pages;
1048 /* number of mmap range requests */
1049 pgoff_t ras_range_requests;
1052 * Optimal RPC size in pages.
1053 * It decides how many pages will be sent for each read-ahead.
1055 unsigned long ras_rpc_pages;
1057 * Where next read-ahead should start at. This lies within read-ahead
1058 * window. Read-ahead window is read in pieces rather than at once
1059 * because: 1. lustre limits total number of pages under read-ahead by
1060 * ->ra_max_pages (see ll_ra_count_get()), 2. client cannot read pages
1061 * not covered by DLM lock.
1063 pgoff_t ras_next_readahead_idx;
1065 * Total number of ll_file_read requests issued, reads originating
1066 * due to mmap are not counted in this total. This value is used to
1067 * trigger full file read-ahead after multiple reads to a small file.
1069 unsigned long ras_requests;
1071 * The following 3 items are used for detecting the stride I/O
1073 * In stride I/O mode,
1074 * ...............|-----data-----|****gap*****|--------|******|....
1075 * offset |-stride_bytes-|-stride_gap-|
1076 * ras_stride_offset = offset;
1077 * ras_stride_length = stride_bytes + stride_gap;
1078 * ras_stride_bytes = stride_bytes;
1079 * Note: all these three items are counted by bytes.
1081 loff_t ras_stride_offset;
1082 loff_t ras_stride_length;
1083 loff_t ras_stride_bytes;
1085 * number of consecutive stride request count, and it is similar as
1086 * ras_consecutive_requests, but used for stride I/O mode.
1087 * Note: only more than 2 consecutive stride request are detected,
1088 * stride read-ahead will be enable
1090 unsigned long ras_consecutive_stride_requests;
1091 /* index of the last page that async readahead starts */
1092 pgoff_t ras_async_last_readpage_idx;
1093 /* whether we should increase readahead window */
1094 bool ras_need_increase_window;
1095 /* we're attempting to read the whole file */
1096 bool ras_whole_file_read;
1099 struct ll_readahead_work {
1100 /** File to readahead */
1101 struct file *lrw_file;
1102 pgoff_t lrw_start_idx;
1103 pgoff_t lrw_end_idx;
1106 /* async worker to handler read */
1107 struct work_struct lrw_readahead_work;
1108 char lrw_jobid[LUSTRE_JOBID_SIZE];
1111 extern struct kmem_cache *ll_file_data_slab;
1112 struct lustre_handle;
1113 struct ll_file_data {
1114 struct ll_readahead_state fd_ras;
1115 struct ll_grouplock fd_grouplock;
1119 /* openhandle if lease exists for this file.
1120 * Borrow lli->lli_och_mutex to protect assignment
1122 struct obd_client_handle *fd_lease_och;
1123 struct obd_client_handle *fd_och;
1124 struct file *fd_file;
1125 /* Indicate whether need to report failure when close.
1126 * true: failure is known, not report again.
1127 * false: unknown failure, should report.
1129 bool fd_write_failed;
1130 bool ll_lock_no_expand;
1131 /* Used by mirrored file to lead IOs to a specific mirror, usually
1132 * for mirror resync. 0 means default.
1134 __u32 fd_designated_mirror;
1135 /* The layout version when resync starts. Resync I/O should carry this
1136 * layout version for verification to OST objects
1138 __u32 fd_layout_version;
1139 struct pcc_file fd_pcc_file;
1140 /* striped directory may read partially if some stripe inaccessible,
1141 * -errno is saved here, and will return to user in close().
1143 int fd_partial_readdir_rc;
1144 /* mdtest unique/shared dir stat mode: per process statahead struct. */
1145 struct ll_statahead_info *fd_sai;
1148 void llite_tunables_unregister(void);
1149 int llite_tunables_register(void);
1151 static inline struct inode *ll_info2i(struct ll_inode_info *lli)
1153 return &lli->lli_vfs_inode;
1156 __u32 ll_i2suppgid(struct inode *i);
1157 void ll_i2gids(__u32 *suppgids, struct inode *i1, struct inode *i2);
1159 static inline int ll_need_32bit_api(struct ll_sb_info *sbi)
1161 #if BITS_PER_LONG == 32
1163 #elif defined(CONFIG_COMPAT)
1164 if (unlikely(test_bit(LL_SBI_32BIT_API, sbi->ll_flags)))
1167 # ifdef CONFIG_X86_X32
1168 /* in_compat_syscall() returns true when called from a kthread
1169 * and CONFIG_X86_X32 is enabled, which is wrong. So check
1170 * whether the caller comes from a syscall (ie. not a kthread)
1171 * before calling in_compat_syscall().
1173 if (current->flags & PF_KTHREAD)
1177 return unlikely(in_compat_syscall());
1179 return unlikely(test_bit(LL_SBI_32BIT_API, sbi->ll_flags));
1183 static inline bool ll_sbi_has_fast_read(struct ll_sb_info *sbi)
1185 return test_bit(LL_SBI_FAST_READ, sbi->ll_flags);
1188 static inline bool ll_sbi_has_tiny_write(struct ll_sb_info *sbi)
1190 return test_bit(LL_SBI_TINY_WRITE, sbi->ll_flags);
1193 static inline bool ll_sbi_has_file_heat(struct ll_sb_info *sbi)
1195 return test_bit(LL_SBI_FILE_HEAT, sbi->ll_flags);
1198 static inline bool ll_sbi_has_foreign_symlink(struct ll_sb_info *sbi)
1200 return test_bit(LL_SBI_FOREIGN_SYMLINK, sbi->ll_flags);
1203 static inline bool ll_sbi_has_parallel_dio(struct ll_sb_info *sbi)
1205 return test_bit(LL_SBI_PARALLEL_DIO, sbi->ll_flags);
1208 static inline bool ll_sbi_has_unaligned_dio(struct ll_sb_info *sbi)
1210 return test_bit(LL_SBI_UNALIGNED_DIO, sbi->ll_flags);
1213 void ll_ras_enter(struct file *f, loff_t pos, size_t bytes);
1215 /* llite/lcommon_misc.c */
1216 int cl_ocd_update(struct obd_device *host, struct obd_device *watched,
1217 enum obd_notify_event ev, void *owner);
1218 int cl_get_grouplock(struct cl_object *obj, unsigned long gid, int nonblock,
1219 struct ll_grouplock *lg);
1220 void cl_put_grouplock(struct ll_grouplock *lg);
1222 /* llite/lproc_llite.c */
1223 int ll_debugfs_register_super(struct super_block *sb, const char *name);
1224 void ll_debugfs_unregister_super(struct super_block *sb);
1225 void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, long count);
1226 void ll_free_rw_stats_info(struct ll_sb_info *sbi);
1229 LPROC_LL_READ_BYTES,
1230 LPROC_LL_WRITE_BYTES,
1257 LPROC_LL_GETXATTR_HITS,
1259 LPROC_LL_REMOVEXATTR,
1260 LPROC_LL_INODE_PERM,
1262 LPROC_LL_INODE_OCOUNT,
1263 LPROC_LL_INODE_OPCLTM,
1264 LPROC_LL_FILE_OPCODES
1268 enum get_default_layout_type {
1269 GET_DEFAULT_LAYOUT_ROOT = 1,
1272 extern const struct file_operations ll_dir_operations;
1273 extern const struct inode_operations ll_dir_inode_operations;
1274 extern struct kmem_cache *quota_iter_slab;
1275 #ifdef HAVE_DIR_CONTEXT
1276 int ll_dir_read(struct inode *inode, __u64 *pos, struct md_op_data *op_data,
1277 struct dir_context *ctx, int *partial_readdir_rc);
1279 int ll_dir_read(struct inode *inode, __u64 *pos, struct md_op_data *op_data,
1280 void *cookie, filldir_t filldir, int *partial_readdir_rc);
1282 int ll_get_mdt_idx(struct inode *inode);
1283 int ll_get_mdt_idx_by_fid(struct ll_sb_info *sbi, const struct lu_fid *fid);
1284 struct page *ll_get_dir_page(struct inode *dir, struct md_op_data *op_data,
1285 __u64 offset, int *partial_readdir_rc);
1286 void ll_release_page(struct inode *inode, struct page *page, bool remove);
1287 int quotactl_ioctl(struct super_block *sb, struct if_quotactl *qctl);
1288 void ll_quota_iter_check_and_cleanup(struct ll_sb_info *sbi, bool check);
1291 extern const struct inode_operations ll_special_inode_operations;
1293 struct inode *ll_iget(struct super_block *sb, ino_t hash,
1294 struct lustre_md *lic);
1295 int ll_test_inode_by_fid(struct inode *inode, void *opaque);
1296 int ll_md_blocking_ast(struct ldlm_lock *lock, struct ldlm_lock_desc *ldesc,
1297 void *data, int flag);
1298 struct dentry *ll_splice_alias(struct inode *inode, struct dentry *de);
1299 int ll_rmdir_entry(struct inode *dir, char *name, int namelen);
1300 void ll_update_times(struct ptlrpc_request *request, struct inode *inode);
1303 int ll_writepage(struct page *page, struct writeback_control *wbc);
1304 int ll_writepages(struct address_space *a, struct writeback_control *wbc);
1305 int ll_readpage(struct file *file, struct page *page);
1306 #ifdef HAVE_AOPS_READ_FOLIO
1307 int ll_read_folio(struct file *file, struct folio *folio);
1309 int ll_io_read_page(const struct lu_env *env, struct cl_io *io,
1310 struct cl_page *page, struct file *file);
1311 void ll_readahead_init(struct inode *inode, struct ll_readahead_state *ras);
1312 int vvp_io_write_commit(const struct lu_env *env, struct cl_io *io);
1315 void ll_cl_add(struct inode *inode, const struct lu_env *env, struct cl_io *io,
1316 enum lcc_type type);
1317 void ll_cl_remove(struct inode *inode, const struct lu_env *env);
1318 struct ll_cl_context *ll_cl_find(struct inode *inode);
1320 extern const struct address_space_operations ll_aops;
1323 extern const struct inode_operations ll_file_inode_operations;
1324 const struct file_operations *ll_select_file_operations(struct ll_sb_info *sbi);
1325 extern int ll_have_md_lock(struct obd_export *exp, struct inode *inode,
1326 __u64 *bits, enum ldlm_mode l_req_mode);
1327 extern enum ldlm_mode ll_take_md_lock(struct inode *inode, __u64 bits,
1328 struct lustre_handle *lockh, __u64 flags,
1329 enum ldlm_mode mode);
1331 int ll_file_open(struct inode *inode, struct file *file);
1332 int ll_file_release(struct inode *inode, struct file *file);
1333 int ll_release_openhandle(struct dentry *d, struct lookup_intent *l);
1334 int ll_md_real_close(struct inode *inode, fmode_t fmode);
1335 void ll_track_file_opens(struct inode *inode);
1336 extern void ll_rw_stats_tally(struct ll_sb_info *sbi, pid_t pid,
1337 struct ll_file_data *file, loff_t pos,
1338 size_t count, int rw);
1339 #if defined(HAVE_USER_NAMESPACE_ARG) || defined(HAVE_INODEOPS_ENHANCED_GETATTR)
1340 int ll_getattr(struct mnt_idmap *, const struct path *path,
1341 struct kstat *stat, u32 request_mask, unsigned int flags);
1343 int ll_getattr(struct vfsmount *mnt, struct dentry *de, struct kstat *stat);
1344 #endif /* HAVE_USER_NAMESPACE_ARG */
1345 int ll_getattr_dentry(struct dentry *de, struct kstat *stat, u32 request_mask,
1346 unsigned int flags, bool foreign);
1347 #ifdef CONFIG_LUSTRE_FS_POSIX_ACL
1348 struct posix_acl *ll_get_acl(
1349 #ifdef HAVE_ACL_WITH_DENTRY
1350 struct mnt_idmap *, struct dentry *, int);
1351 #elif defined HAVE_GET_ACL_RCU_ARG
1352 struct inode *inode, int type, bool rcu);
1354 struct inode *inode, int type);
1355 #endif /* HAVE_GET_ACL_RCU_ARG */
1357 int ll_set_acl(struct mnt_idmap *mnt_userns,
1358 #ifdef HAVE_ACL_WITH_DENTRY
1359 struct dentry *dentry,
1361 struct inode *inode,
1363 struct posix_acl *acl, int type);
1364 #else /* !CONFIG_LUSTRE_FS_POSIX_ACL */
1365 #define ll_get_acl NULL
1366 #define ll_set_acl NULL
1367 #endif /* CONFIG_LUSTRE_FS_POSIX_ACL */
1369 static inline int ll_xflags_to_inode_flags(int xflags)
1371 return ((xflags & FS_XFLAG_SYNC) ? S_SYNC : 0) |
1372 ((xflags & FS_XFLAG_NOATIME) ? S_NOATIME : 0) |
1373 ((xflags & FS_XFLAG_APPEND) ? S_APPEND : 0) |
1374 ((xflags & FS_XFLAG_IMMUTABLE) ? S_IMMUTABLE : 0);
1377 static inline int ll_inode_flags_to_xflags(int inode_flags)
1379 return ((inode_flags & S_SYNC) ? FS_XFLAG_SYNC : 0) |
1380 ((inode_flags & S_NOATIME) ? FS_XFLAG_NOATIME : 0) |
1381 ((inode_flags & S_APPEND) ? FS_XFLAG_APPEND : 0) |
1382 ((inode_flags & S_IMMUTABLE) ? FS_XFLAG_IMMUTABLE : 0);
1385 int ll_migrate(struct inode *parent, struct file *file,
1386 struct lmv_user_md *lum, const char *name, __u32 flags);
1387 int ll_get_fid_by_name(struct inode *parent, const char *name,
1388 int namelen, struct lu_fid *fid, struct inode **inode);
1389 int ll_inode_permission(struct mnt_idmap *mnt, struct inode *inode, int mask);
1390 int ll_ioctl_check_project(struct inode *inode, __u32 xflags, __u32 projid);
1391 int ll_set_project(struct inode *inode, __u32 xflags, __u32 projid);
1392 #ifndef HAVE_FILEATTR_GET
1393 int ll_ioctl_fsgetxattr(struct inode *inode, unsigned int cmd,
1395 int ll_ioctl_fssetxattr(struct inode *inode, unsigned int cmd,
1398 int ll_ioctl_project(struct file *file, unsigned int cmd, void __user *uarg);
1399 int ll_ioctl_ahead(struct file *file, struct llapi_lu_ladvise2 *ladvise);
1401 int ll_lov_setstripe_ea_info(struct inode *inode, struct dentry *dentry,
1402 __u64 flags, struct lov_user_md *lum,
1404 int ll_lov_getstripe_ea_info(struct inode *inode, const char *filename,
1405 struct lov_mds_md **lmm, int *lmm_size,
1406 struct ptlrpc_request **request);
1407 int ll_dir_setstripe(struct inode *inode, struct lov_user_md *lump,
1409 int ll_dir_get_default_layout(struct inode *inode, void **plmm, int *plmm_size,
1410 struct ptlrpc_request **request, u64 valid,
1411 enum get_default_layout_type type);
1412 int ll_dir_getstripe_default(struct inode *inode, void **lmmp,
1413 int *lmm_size, struct ptlrpc_request **request,
1414 struct ptlrpc_request **root_request, u64 valid);
1415 int ll_dir_getstripe(struct inode *inode, void **plmm, int *plmm_size,
1416 struct ptlrpc_request **request, u64 valid);
1417 int ll_fsync(struct file *file, loff_t start, loff_t end, int data);
1418 int ll_merge_attr(const struct lu_env *env, struct inode *inode);
1419 int ll_merge_attr_try(const struct lu_env *env, struct inode *inode);
1420 int ll_fid2path(struct inode *inode, void __user *arg);
1421 int __ll_fid2path(struct inode *inode, struct getinfo_fid2path *gfout,
1422 size_t outsize, __u32 pathlen_orig);
1423 int ll_data_version(struct inode *inode, __u64 *data_version, int flags);
1424 int ll_hsm_release(struct inode *inode);
1425 int ll_hsm_state_set(struct inode *inode, struct hsm_state_set *hss);
1426 void ll_io_set_mirror(struct cl_io *io, const struct file *file);
1428 /* llite/dcache.c */
1430 extern const struct dentry_operations ll_d_ops;
1432 bool ll_d_setup(struct dentry *de, bool do_put);
1434 static inline bool lld_is_init(struct dentry *dentry)
1436 return ll_d2d(dentry);
1439 #define ll_d_setup(de, do_put) (true)
1440 #define lld_is_init(dentry) (true)
1443 void ll_intent_drop_lock(struct lookup_intent *lookup);
1444 void ll_intent_release(struct lookup_intent *lookup);
1445 void ll_prune_aliases(struct inode *inode);
1446 void ll_lookup_finish_locks(struct lookup_intent *it, struct dentry *dentry);
1447 int ll_revalidate_it_finish(struct ptlrpc_request *request,
1448 struct lookup_intent *it, struct dentry *de);
1450 /* llite/llite_lib.c */
1451 extern const struct super_operations lustre_super_operations;
1453 void ll_lli_init(struct ll_inode_info *lli);
1454 int ll_fill_super(struct super_block *sb);
1455 void ll_put_super(struct super_block *sb);
1456 void ll_kill_super(struct super_block *sb);
1457 struct inode *ll_inode_from_resource_lock(struct ldlm_lock *lock);
1458 void ll_dir_clear_lsm_md(struct inode *inode);
1459 void ll_clear_inode(struct inode *inode);
1460 int volatile_ref_file(const char *volatile_name, int volatile_len,
1461 struct file **ref_file);
1462 int ll_setattr_raw(struct dentry *dentry, struct iattr *attr,
1463 enum op_xvalid xvalid, bool hsm_import);
1464 int ll_setattr(struct mnt_idmap *map, struct dentry *de, struct iattr *attr);
1465 int ll_statfs(struct dentry *de, struct kstatfs *sfs);
1466 int ll_statfs_internal(struct ll_sb_info *sbi, struct obd_statfs *osfs,
1468 int ll_update_inode(struct inode *inode, struct lustre_md *md);
1469 void ll_update_inode_flags(struct inode *inode, unsigned int ext_flags);
1470 void ll_update_dir_depth_dmv(struct inode *dir, struct dentry *de);
1471 int ll_read_inode2(struct inode *inode, void *opaque);
1472 void ll_truncate_inode_pages_final(struct inode *inode);
1473 void ll_delete_inode(struct inode *inode);
1474 #ifdef HAVE_FILEATTR_GET
1475 int ll_fileattr_get(struct dentry *dentry, struct fileattr *fa);
1476 int ll_fileattr_set(struct mnt_idmap *mnt_userns,
1477 struct dentry *dentry, struct fileattr *fa);
1479 int ll_iocontrol(struct inode *inode, struct file *file,
1480 unsigned int cmd, void __user *uarg);
1481 int ll_flush_ctx(struct inode *inode);
1482 void ll_umount_begin(struct super_block *sb);
1483 int ll_remount_fs(struct super_block *sb, int *flags, char *data);
1484 int ll_show_options(struct seq_file *seq, struct dentry *dentry);
1485 void ll_dirty_page_discard_warn(struct inode *inode, int ioret);
1486 int ll_prep_inode(struct inode **inode, struct req_capsule *pill,
1487 struct super_block *sb, struct lookup_intent *it);
1488 int ll_obd_statfs(struct inode *inode, void __user *arg);
1489 int ll_get_max_mdsize(struct ll_sb_info *sbi, int *max_mdsize);
1490 int ll_get_default_mdsize(struct ll_sb_info *sbi, int *default_mdsize);
1491 int ll_set_default_mdsize(struct ll_sb_info *sbi, int default_mdsize);
1493 void ll_unlock_md_op_lsm(struct md_op_data *op_data);
1494 struct md_op_data *ll_prep_md_op_data(struct md_op_data *op_data,
1495 struct inode *i1, struct inode *i2,
1496 const char *name, size_t namelen,
1497 __u32 mode, enum md_op_code opc,
1499 void ll_finish_md_op_data(struct md_op_data *op_data);
1500 int ll_get_obd_name(struct inode *inode, unsigned int cmd, void __user *arg);
1501 void ll_compute_rootsquash_state(struct ll_sb_info *sbi);
1502 ssize_t ll_copy_user_md(const struct lov_user_md __user *md,
1503 struct lov_user_md **kbuf);
1504 void ll_open_cleanup(struct super_block *sb, struct req_capsule *pill);
1506 void ll_dom_finish_open(struct inode *inode, struct ptlrpc_request *req);
1508 /* Compute expected user md size when passing in a md from user space */
1509 static inline ssize_t ll_lov_user_md_size(const struct lov_user_md *lum)
1511 switch (lum->lmm_magic) {
1512 case LOV_USER_MAGIC_V1:
1513 return sizeof(struct lov_user_md_v1);
1514 case LOV_USER_MAGIC_V3:
1515 return sizeof(struct lov_user_md_v3);
1516 case LOV_USER_MAGIC_SPECIFIC:
1517 if (lum->lmm_stripe_count > LOV_MAX_STRIPE_COUNT)
1520 return lov_user_md_size(lum->lmm_stripe_count,
1521 LOV_USER_MAGIC_SPECIFIC);
1522 case LOV_USER_MAGIC_COMP_V1:
1523 return ((struct lov_comp_md_v1 *)lum)->lcm_size;
1524 case LOV_USER_MAGIC_FOREIGN:
1525 return lov_foreign_size(lum);
1531 /* llite/llite_nfs.c */
1532 extern const struct export_operations lustre_export_operations;
1533 __u32 get_uuid2int(const char *name, int len);
1534 struct inode *search_inode_for_lustre(struct super_block *sb,
1535 const struct lu_fid *fid);
1536 int ll_dir_get_parent_fid(struct inode *dir, struct lu_fid *parent_fid);
1538 /* llite/symlink.c */
1539 extern const struct inode_operations ll_fast_symlink_inode_operations;
1541 /* IO arguments for various VFS I/O interfaces. */
1542 struct vvp_io_args {
1543 /** normal/sendfile/splice */
1546 struct kiocb *via_iocb;
1547 struct iov_iter *via_iter;
1550 /* did we switch this IO from BIO to DIO using hybrid IO? */
1551 int via_hybrid_switched:1;
1554 static inline unsigned int iocb_ki_flags_get(const struct file *file,
1555 const struct kiocb *iocb)
1558 return iocb ? iocb->ki_flags : 0;
1560 return file->f_flags;
1564 static inline unsigned int vvp_io_args_flags(const struct file *file,
1565 const struct vvp_io_args *args)
1567 return iocb_ki_flags_get(file, args ? args->u.normal.via_iocb : NULL);
1575 struct ll_cl_context {
1576 struct list_head lcc_list;
1578 const struct lu_env *lcc_env;
1579 struct cl_io *lcc_io;
1580 struct cl_page *lcc_page;
1581 enum lcc_type lcc_type;
1582 pgoff_t lcc_end_index;
1585 struct ll_thread_info {
1586 struct vvp_io_args lti_args;
1587 struct ra_io_arg lti_ria;
1588 struct ll_cl_context lti_io_ctx;
1591 extern struct lu_context_key ll_thread_key;
1593 static inline struct ll_thread_info *ll_env_info(const struct lu_env *env)
1595 struct ll_thread_info *lti;
1597 lti = lu_context_key_get(&env->le_ctx, &ll_thread_key);
1598 LASSERT(lti != NULL);
1603 static inline struct vvp_io_args *ll_env_args(const struct lu_env *env)
1605 return &ll_env_info(env)->lti_args;
1608 void ll_io_init(struct cl_io *io, struct file *file, enum cl_io_type iot,
1609 struct vvp_io_args *args);
1611 /* llite/llite_mmap.c */
1613 int ll_file_mmap(struct file *file, struct vm_area_struct *vma);
1614 void policy_from_vma(union ldlm_policy_data *policy, struct vm_area_struct *vma,
1615 unsigned long addr, size_t count);
1616 struct vm_area_struct *our_vma(struct mm_struct *mm, unsigned long addr,
1619 #define ll_s2sbi(sb) (s2lsi(sb)->lsi_llsbi)
1621 /* don't need an addref as the sb_info should be holding one */
1622 static inline struct obd_export *ll_s2dtexp(struct super_block *sb)
1624 return ll_s2sbi(sb)->ll_dt_exp;
1627 /* don't need an addref as the sb_info should be holding one */
1628 static inline struct obd_export *ll_s2mdexp(struct super_block *sb)
1630 return ll_s2sbi(sb)->ll_md_exp;
1633 static inline struct client_obd *sbi2mdc(struct ll_sb_info *sbi)
1635 struct obd_device *obd = sbi->ll_md_exp->exp_obd;
1642 // FIXME: replace the name of this with LL_SB to conform to kernel stuff
1643 static inline struct ll_sb_info *ll_i2sbi(struct inode *inode)
1645 return ll_s2sbi(inode->i_sb);
1648 static inline struct obd_export *ll_i2dtexp(struct inode *inode)
1650 return ll_s2dtexp(inode->i_sb);
1653 static inline struct obd_export *ll_i2mdexp(struct inode *inode)
1655 return ll_s2mdexp(inode->i_sb);
1658 static inline struct lu_fid *ll_inode2fid(struct inode *inode)
1662 LASSERT(inode != NULL);
1663 fid = &ll_i2info(inode)->lli_fid;
1668 static inline bool ll_dir_striped_locked(struct inode *inode)
1674 if (!S_ISDIR(inode->i_mode))
1677 rc = !!(ll_i2info(inode)->lli_lsm_obj &&
1678 lmv_dir_striped(ll_i2info(inode)->lli_lsm_obj));
1683 static inline bool ll_dir_striped(struct inode *inode)
1687 if (!S_ISDIR(inode->i_mode))
1690 down_read(&ll_i2info(inode)->lli_lsm_sem);
1691 rc = ll_dir_striped_locked(inode);
1692 up_read(&ll_i2info(inode)->lli_lsm_sem);
1697 static inline loff_t ll_file_maxbytes(struct inode *inode)
1699 struct cl_object *obj = ll_i2info(inode)->lli_clob;
1702 return MAX_LFS_FILESIZE;
1704 return min_t(loff_t, cl_object_maxbytes(obj), MAX_LFS_FILESIZE);
1708 extern const struct xattr_handler *ll_xattr_handlers[];
1710 #define XATTR_USER_T 1
1711 #define XATTR_TRUSTED_T 2
1712 #define XATTR_SECURITY_T 3
1713 #define XATTR_ACL_ACCESS_T 4
1714 #define XATTR_ACL_DEFAULT_T 5
1715 #define XATTR_LUSTRE_T 6
1716 #define XATTR_OTHER_T 7
1717 #define XATTR_ENCRYPTION_T 9
1719 ssize_t ll_listxattr(struct dentry *dentry, char *buffer, size_t size);
1720 int ll_xattr_list(struct inode *inode, const char *name, int type,
1721 void *buffer, size_t size, u64 valid);
1722 const struct xattr_handler *get_xattr_type(const char *name);
1724 /* Common IO arguments for various VFS I/O interfaces. */
1725 int cl_sb_init(struct super_block *sb);
1726 int cl_sb_fini(struct super_block *sb);
1728 enum ras_update_flags {
1732 void ll_ra_count_put(struct ll_sb_info *sbi, unsigned long len);
1733 void ll_ra_stats_inc(struct inode *inode, enum ra_stat which);
1737 #define LL_SA_REQ_MIN 2
1738 #define LL_SA_REQ_MIN_DEF 8
1739 #define LL_SA_REQ_MAX 2048
1740 #define LL_SA_REQ_MAX_DEF 128
1742 #define LL_SA_TIMEOUT_DEF 30
1744 /* XXX: To support more concurrent statahead instances, please consider to
1745 * decentralize the RPC lists attached on related import, such as
1746 * imp_{sending,delayed}_list. LU-11079
1748 #define LL_SA_RUNNING_MAX 256
1749 #define LL_SA_RUNNING_DEF 16
1751 #define LL_SA_BATCH_MAX 1024
1752 #define LL_SA_BATCH_DEF 64
1754 #define LL_SA_CACHE_BIT 6
1755 #define LL_SA_CACHE_SIZE (1 << LL_SA_CACHE_BIT)
1756 #define LL_SA_CACHE_MASK (LL_SA_CACHE_SIZE - 1)
1758 #define LSA_FN_PREDICT_HIT 2
1759 #define LSA_FN_MATCH_HIT 4
1761 /* statahead controller, per process struct, for dir only */
1762 struct ll_statahead_info {
1764 struct list_head sai_item;
1765 struct dentry *sai_dentry;
1766 atomic_t sai_refcount; /* On access, hold refcount */
1767 atomic_t sai_inuse_count;/* inuse entry count */
1768 unsigned int sai_max; /* max ahead of lookup */
1769 __u64 sai_sent; /* stat requests sent count */
1770 __u64 sai_replied; /* stat req which recv reply */
1771 __u64 sai_index; /* index of statahead entry */
1772 /* idx entry which is the caller is waiting for */
1773 __u64 sai_index_wait;
1774 __u64 sai_hit; /* hit count */
1775 __u64 sai_miss; /* miss count:
1776 * for "ls -al" case, includes
1777 * hidden dentry miss;
1778 * for "ls -l" case, it does not
1779 * include hidden dentry miss.
1780 * "sai_miss_hidden" is used for
1783 unsigned int sai_consecutive_miss; /* consecutive miss */
1784 /* "ls -al", but first dentry is not a hidden one */
1785 unsigned int sai_miss_hidden;
1786 unsigned int sai_skip_hidden;/* skipped hidden dentry cnt */
1787 /* "ls -al", do stat-ahead for hidden entries */
1788 unsigned int sai_ls_all:1;
1789 wait_queue_head_t sai_waitq; /* stat-ahead wait queue */
1790 struct task_struct *sai_task; /* stat-ahead thread */
1791 struct task_struct *sai_agl_task; /* AGL thread */
1792 struct list_head sai_entries; /* completed entries */
1793 struct list_head sai_agls; /* AGLs to be sent */
1794 atomic_t sai_cache_count; /* entry count in cache */
1795 struct lu_batch *sai_bh;
1796 __u32 sai_max_batch_count;
1797 __u64 sai_index_end;
1800 /* for ADVISE statahead pattern */
1806 /* for FNAME statahead pattern */
1808 __u64 sai_fname_index;
1810 * The length of file name statahead pattern where the
1811 * front part is padding with 0.
1813 __u8 sai_fname_zeroed_len;
1816 char sai_fname[NAME_MAX];
1819 /* Per inode statahead information */
1820 struct ll_statahead_context {
1821 struct inode *sax_inode;
1822 atomic_t sax_refcount;
1824 struct list_head sax_cache[LL_SA_CACHE_SIZE];
1825 spinlock_t sax_cache_lock[LL_SA_CACHE_SIZE];
1826 struct list_head sax_sai_list;
1829 int ll_revalidate_statahead(struct inode *dir, struct dentry **dentry,
1831 int ll_start_statahead(struct inode *dir, struct dentry *dentry, bool agl);
1832 void ll_authorize_statahead(struct inode *dir, void *key);
1833 void ll_deauthorize_statahead(struct inode *dir, void *key);
1834 void ll_statahead_enter(struct inode *dir, struct dentry *dentry);
1837 blkcnt_t dirty_cnt(struct inode *inode);
1839 int cl_glimpse_size0(struct inode *inode, int agl);
1840 int cl_glimpse_lock(const struct lu_env *env, struct cl_io *io,
1841 struct inode *inode, struct cl_object *clob, int agl);
1843 static inline int cl_glimpse_size(struct inode *inode)
1845 return cl_glimpse_size0(inode, 0);
1848 /* AGL is 'async glimpse lock',is speculative lock taken as part of statahead */
1849 static inline int cl_agl(struct inode *inode)
1851 return cl_glimpse_size0(inode, 1);
1854 int ll_file_lock_ahead(struct file *file, struct llapi_lu_ladvise *ladvise);
1856 int cl_io_get(struct inode *inode, struct lu_env **envout,
1857 struct cl_io **ioout, __u16 *refcheck);
1859 static inline int ll_glimpse_size(struct inode *inode)
1861 struct ll_inode_info *lli = ll_i2info(inode);
1864 down_read(&lli->lli_glimpse_sem);
1865 rc = cl_glimpse_size(inode);
1866 lli->lli_glimpse_time = ktime_get();
1867 up_read(&lli->lli_glimpse_sem);
1871 /* dentry may statahead when statahead is enabled and current process has opened
1872 * parent directory, and this dentry hasn't accessed statahead cache before
1875 dentry_may_statahead(struct inode *dir, struct dentry *dentry)
1877 struct ll_inode_info *lli;
1878 struct ll_dentry_data *ldd;
1880 if (ll_i2sbi(dir)->ll_sa_max == 0)
1883 lli = ll_i2info(dir);
1885 /* statahead is not allowed for this dir, there may be three causes:
1886 * 1. dir is not opened.
1887 * 2. statahead hit ratio is too low.
1888 * 3. previous stat started statahead thread failed.
1890 if (!lli->lli_sa_enabled)
1894 * When stating a dentry, kernel may trigger 'revalidate' or 'lookup'
1895 * multiple times, eg. for 'getattr', 'getxattr' and etc.
1896 * For patchless client, lookup intent is not accurate, which may
1897 * misguide statahead. For example:
1898 * The 'revalidate' call for 'getattr' and 'getxattr' of a dentry will
1899 * have the same intent -- IT_GETATTR, while one dentry should access
1900 * statahead cache once, otherwise statahead windows is messed up.
1901 * The solution is as following:
1902 * Assign 'lld_sa_generation' with 'lli_sa_generation' when a dentry
1903 * IT_GETATTR for the first time, and subsequent IT_GETATTR will
1904 * bypass interacting with statahead cache by checking
1905 * 'lld_sa_generation == lli->lli_sa_generation'.
1907 ldd = ll_d2d(dentry);
1908 if (ldd != NULL && lli->lli_sa_generation &&
1909 ldd->lld_sa_generation == lli->lli_sa_generation)
1912 if (lli->lli_sa_pattern & LSA_PATTERN_ADVISE)
1915 if (lli->lli_sa_pattern & (LSA_PATTERN_FNAME | LSA_PATTERN_FN_PREDICT))
1918 /* not the same process, don't statahead */
1919 if (lli->lli_stat_pid != current->pid)
1925 int cl_sync_file_range(struct inode *inode, loff_t start, loff_t end,
1926 enum cl_fsync_mode mode, int ignore_layout);
1928 static inline int ll_file_nolock(const struct file *file)
1930 struct ll_file_data *fd = file->private_data;
1931 struct inode *inode = file_inode((struct file *)file);
1933 LASSERT(fd != NULL);
1934 return ((fd->fd_flags & LL_FILE_IGNORE_LOCK) ||
1935 test_bit(LL_SBI_NOLCK, ll_i2sbi(inode)->ll_flags));
1938 static inline void ll_set_lock_data(struct obd_export *exp, struct inode *inode,
1939 struct lookup_intent *it, __u64 *bits)
1941 if (!it->it_lock_set) {
1942 struct lustre_handle handle;
1944 /* If this inode is a remote object, it will get two
1945 * separate locks in different namespaces, Master MDT,
1946 * where the name entry is, will grant LOOKUP lock,
1947 * remote MDT, where the object is, will grant
1948 * UPDATE|PERM lock. The inode will be attched to both
1949 * LOOKUP and PERM locks, so revoking either locks will
1950 * case the dcache being cleared
1952 if (it->it_remote_lock_mode) {
1953 handle.cookie = it->it_remote_lock_handle;
1954 CDEBUG(D_DLMTRACE, "setting l_data to inode "DFID
1955 "(%p) for remote lock %#llx\n",
1956 PFID(ll_inode2fid(inode)), inode,
1958 md_set_lock_data(exp, &handle, inode, NULL);
1961 handle.cookie = it->it_lock_handle;
1964 "setting l_data to inode "DFID"(%p) for lock %#llx\n",
1965 PFID(ll_inode2fid(inode)), inode, handle.cookie);
1967 md_set_lock_data(exp, &handle, inode, &it->it_lock_bits);
1968 it->it_lock_set = 1;
1972 *bits = it->it_lock_bits;
1975 static inline int d_lustre_invalid(const struct dentry *dentry)
1977 return !ll_d2d(dentry) || ll_d2d(dentry)->lld_invalid;
1981 * Mark dentry INVALID, if dentry refcount is zero (this is normally case for
1982 * ll_md_blocking_ast), it will be pruned by ll_prune_aliases() and
1983 * ll_prune_negative_children(); otherwise dput() of the last refcount will
1984 * unhash this dentry and kill it.
1986 static inline void d_lustre_invalidate(struct dentry *dentry)
1988 CDEBUG(D_DENTRY, "invalidate dentry %pd (%p) parent %p inode %p refc %d\n",
1990 dentry->d_parent, dentry->d_inode, ll_d_count(dentry));
1992 spin_lock(&dentry->d_lock);
1993 if (lld_is_init(dentry))
1994 ll_d2d(dentry)->lld_invalid = 1;
1995 spin_unlock(&dentry->d_lock);
1998 static inline void d_lustre_revalidate(struct dentry *dentry)
2000 spin_lock(&dentry->d_lock);
2001 LASSERT(ll_d2d(dentry));
2002 ll_d2d(dentry)->lld_invalid = 0;
2003 spin_unlock(&dentry->d_lock);
2006 static inline dev_t ll_compat_encode_dev(dev_t dev)
2008 /* The compat_sys_*stat*() syscalls will fail unless the
2009 * device majors and minors are both less than 256. Note that
2010 * the value returned here will be passed through
2011 * old_encode_dev() in cp_compat_stat(). And so we are not
2012 * trying to return a valid compat (u16) device number, just
2013 * one that will pass the old_valid_dev() check.
2016 return MKDEV(MAJOR(dev) & 0xff, MINOR(dev) & 0xff);
2019 int ll_layout_conf(struct inode *inode, const struct cl_object_conf *conf);
2020 int ll_layout_refresh(struct inode *inode, __u32 *gen);
2021 int ll_layout_restore(struct inode *inode, loff_t start, __u64 length);
2022 int ll_layout_write_intent(struct inode *inode, enum layout_intent_opc opc,
2023 struct lu_extent *ext);
2025 int ll_xattr_init(void);
2026 void ll_xattr_fini(void);
2028 int ll_page_sync_io(const struct lu_env *env, struct cl_io *io,
2029 struct cl_page *page, enum cl_req_type crt);
2031 int ll_getparent(struct file *file, struct getparent __user *arg);
2034 int cl_setattr_ost(struct cl_object *obj, const struct iattr *attr,
2035 enum op_xvalid xvalid, unsigned int attr_flags);
2037 extern struct lu_env *cl_inode_fini_env;
2038 extern __u16 cl_inode_fini_refcheck;
2040 int cl_file_inode_init(struct inode *inode, struct lustre_md *md);
2041 void cl_inode_fini(struct inode *inode);
2043 u64 cl_fid_build_ino(const struct lu_fid *fid, int api32);
2044 u32 cl_fid_build_gen(const struct lu_fid *fid);
2046 static inline struct pcc_super *ll_i2pccs(struct inode *inode)
2048 return &ll_i2sbi(inode)->ll_pcc_super;
2051 static inline struct pcc_super *ll_info2pccs(struct ll_inode_info *lli)
2053 return ll_i2pccs(ll_info2i(lli));
2057 /* The digested form is made of a FID (16 bytes) followed by the second-to-last
2058 * ciphertext block (16 bytes), so a total length of 32 bytes.
2059 * That way, llcrypt does not compute a digested form of this digest.
2061 struct ll_digest_filename {
2062 struct lu_fid ldf_fid;
2063 char ldf_excerpt[LL_CRYPTO_BLOCK_SIZE];
2066 int ll_prepare_lookup(struct inode *dir, struct dentry *de,
2067 struct llcrypt_name *fname, struct lu_fid *fid);
2068 int ll_setup_filename(struct inode *dir, const struct qstr *iname,
2069 int lookup, struct llcrypt_name *fname,
2070 struct lu_fid *fid);
2071 int ll_fname_disk_to_usr(struct inode *inode,
2072 u32 hash, u32 minor_hash,
2073 struct llcrypt_str *iname, struct llcrypt_str *oname,
2074 struct lu_fid *fid);
2075 int ll_file_open_encrypt(struct inode *inode, struct file *filp);
2077 static inline char *xattr_for_enc(struct inode *inode)
2079 if (ll_sbi_has_name_encrypt(ll_i2sbi(inode)))
2080 return LL_XATTR_NAME_ENCRYPTION_CONTEXT;
2082 return LL_XATTR_NAME_ENCRYPTION_CONTEXT_OLD;
2085 #ifdef HAVE_LUSTRE_CRYPTO
2086 extern const struct llcrypt_operations lustre_cryptops;
2089 /* llite/llite_foreign.c */
2090 int ll_manage_foreign(struct inode *inode, struct lustre_md *lmd);
2091 bool ll_foreign_is_openable(struct dentry *dentry, unsigned int flags);
2092 bool ll_foreign_is_removable(struct dentry *dentry, bool unset);
2094 int ll_filemap_fault(struct vm_area_struct *vma, struct vm_fault *vmf);
2096 #endif /* LLITE_INTERNAL_H */