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 */
65 #ifndef LOOKUP_CONTINUE
66 #define LOOKUP_CONTINUE LOOKUP_PARENT
69 #ifdef HAVE_DENTRY_D_U_D_ALIAS
70 #define d_alias d_u.d_alias
73 /** Only used on client-side for indicating the tail of dir hash/offset. */
74 #define LL_DIR_END_OFF 0x7fffffffffffffffULL
75 #define LL_DIR_END_OFF_32BIT 0x7fffffffUL
77 /* 4UL * 1024 * 1024 */
78 #define LL_MAX_BLKSIZE_BITS 22
80 #define LL_IT2STR(it) ((it) ? ldlm_it2str((it)->it_op) : "0")
82 #define TIMES_SET_FLAGS (ATTR_MTIME_SET | ATTR_ATIME_SET | ATTR_TIMES_SET)
84 struct ll_dentry_data {
85 unsigned int lld_sa_generation;
86 unsigned int lld_invalid:1;
87 struct rcu_head lld_rcu_head;
90 #define ll_d2d(de) ((struct ll_dentry_data*)((de)->d_fsdata))
92 #define LLI_INODE_MAGIC 0x111d0de5
93 #define LLI_INODE_DEAD 0xdeadd00d
95 struct ll_getname_data {
96 #ifdef HAVE_DIR_CONTEXT
97 struct dir_context ctx;
99 char *lgd_name; /* points to a buffer with NAME_MAX+1 size */
100 struct lu_fid lgd_fid; /* target fid we are looking for */
101 int lgd_found; /* inode matched? */
104 struct ll_grouplock {
105 struct lu_env *lg_env;
107 struct cl_lock *lg_lock;
108 unsigned long lg_gid;
111 /* See comment on trunc_sem_down_read_nowait */
112 struct ll_trunc_sem {
113 /* when positive, this is a count of readers, when -1, it indicates
114 * the semaphore is held for write, and 0 is unlocked
116 atomic_t ll_trunc_readers;
117 /* this tracks a count of waiting writers */
118 atomic_t ll_trunc_waiters;
121 struct ll_inode_info {
122 __u32 lli_inode_magic;
125 volatile unsigned long lli_flags;
126 struct posix_acl *lli_posix_acl;
128 /* identifying fields for both metadata and data stacks. */
129 struct lu_fid lli_fid;
130 /* master inode fid for stripe directory */
131 struct lu_fid lli_pfid;
133 /* We need all three because every inode may be opened in different
135 struct obd_client_handle *lli_mds_read_och;
136 struct obd_client_handle *lli_mds_write_och;
137 struct obd_client_handle *lli_mds_exec_och;
138 __u64 lli_open_fd_read_count;
139 __u64 lli_open_fd_write_count;
140 __u64 lli_open_fd_exec_count;
142 /* Protects access to och pointers and their usage counters */
143 struct mutex lli_och_mutex;
145 struct inode lli_vfs_inode;
147 /* the most recent timestamps obtained from mds */
152 spinlock_t lli_agl_lock;
153 bool lli_synced_to_mds;
155 /* inode specific open lock caching threshold */
156 u32 lli_open_thrsh_count;
157 /* Number of times this inode was opened */
158 u64 lli_open_fd_count;
159 /* When last close was performed on this inode */
160 ktime_t lli_close_fd_time;
162 /* Try to make the d::member and f::member are aligned. Before using
163 * these members, make clear whether it is directory or not. */
167 /* metadata statahead */
168 /* since parent-child threads can share the same @file
169 * struct, "opendir_key" is the token when dir close for
170 * case of parent exit before child -- it is me should
171 * cleanup the dir readahead. */
172 void *lli_opendir_key;
173 struct ll_statahead_info *lli_sai;
174 /* protect statahead stuff. */
175 spinlock_t lli_sa_lock;
176 /* "opendir_pid" is the token when lookup/revalid
177 * -- I am the owner of dir statahead. */
178 pid_t lli_opendir_pid;
179 /* directory depth to ROOT */
180 unsigned short lli_dir_depth;
181 /* directory depth to ancestor whose default LMV is
184 unsigned short lli_inherit_depth;
185 /* stat will try to access statahead entries or start
186 * statahead if this flag is set, and this flag will be
187 * set upon dir open, and cleared when dir is closed,
188 * statahead hit ratio is too low, or start statahead
190 unsigned short lli_sa_enabled:1;
191 /* generation for statahead */
192 unsigned int lli_sa_generation;
193 /* rw lock protects lli_lsm_md */
194 struct rw_semaphore lli_lsm_sem;
195 /* directory stripe information */
196 struct lmv_stripe_md *lli_lsm_md;
197 /* directory default LMV */
198 struct lmv_stripe_md *lli_default_lsm_md;
201 /* for non-directory */
203 struct mutex lli_size_mutex;
204 char *lli_symlink_name;
205 struct ll_trunc_sem lli_trunc_sem;
206 struct range_lock_tree lli_write_tree;
207 struct mutex lli_setattr_mutex;
209 struct rw_semaphore lli_glimpse_sem;
210 ktime_t lli_glimpse_time;
211 struct list_head lli_agl_list;
214 /* for writepage() only to communicate to fsync */
217 /* protect the file heat fields */
218 spinlock_t lli_heat_lock;
219 __u32 lli_heat_flags;
220 struct obd_heat_instance lli_heat_instances[OBD_HEAT_COUNT];
223 * Whenever a process try to read/write the file, the
224 * jobid of the process will be saved here, and it'll
225 * be packed into the write PRC when flush later.
227 * So the read/write statistics for jobid will not be
228 * accurate if the file is shared by different jobs.
230 char lli_jobid[LUSTRE_JOBID_SIZE];
232 struct mutex lli_pcc_lock;
233 enum lu_pcc_state_flags lli_pcc_state;
235 * @lli_pcc_generation saves the gobal PCC generation
236 * when the file was successfully attached into PCC.
237 * The flags of the PCC dataset are saved in
239 * The gobal PCC generation will be increased when add
240 * or delete a PCC backend, or change the configuration
241 * parameters for PCC.
242 * If @lli_pcc_generation is same as the gobal PCC
243 * generation, we can use the saved flags of the PCC
244 * dataset to determine whether need to try auto attach
247 __u64 lli_pcc_generation;
248 enum pcc_dataset_flags lli_pcc_dsflags;
249 struct pcc_inode *lli_pcc_inode;
251 struct mutex lli_group_mutex;
252 __u64 lli_group_users;
253 unsigned long lli_group_gid;
255 __u64 lli_attr_valid;
257 __u64 lli_lazyblocks;
261 /* XXX: For following frequent used members, although they maybe special
262 * used for non-directory object, it is some time-wasting to check
263 * whether the object is directory or not before using them. On the
264 * other hand, currently, sizeof(f) > sizeof(d), it cannot reduce
265 * the "ll_inode_info" size even if moving those members into u.f.
266 * So keep them out side.
268 * In the future, if more members are added only for directory,
269 * some of the following members can be moved into u.f.
271 struct cl_object *lli_clob;
273 /* mutex to request for layout lock exclusively. */
274 struct mutex lli_layout_mutex;
275 /* Layout version, protected by lli_layout_lock */
276 __u32 lli_layout_gen;
277 spinlock_t lli_layout_lock;
279 __u32 lli_projid; /* project id */
281 struct rw_semaphore lli_xattrs_list_rwsem;
282 struct mutex lli_xattrs_enq_lock;
283 struct list_head lli_xattrs; /* ll_xattr_entry->xe_list */
284 struct list_head lli_lccs; /* list of ll_cl_context */
285 seqlock_t lli_page_inv_lock;
288 #ifndef HAVE_USER_NAMESPACE_ARG
289 #define inode_permission(ns, inode, mask) inode_permission(inode, mask)
290 #define generic_permission(ns, inode, mask) generic_permission(inode, mask)
291 #define simple_setattr(ns, de, iattr) simple_setattr(de, iattr)
292 #define ll_inode_permission(ns, inode, mask) ll_inode_permission(inode, mask)
293 #ifdef HAVE_INODEOPS_ENHANCED_GETATTR
294 #define ll_getattr(ns, path, stat, mask, fl) ll_getattr(path, stat, mask, fl)
295 #endif /* HAVE_INODEOPS_ENHANCED_GETATTR */
296 #define ll_setattr(ns, de, attr) ll_setattr(de, attr)
299 static inline void ll_trunc_sem_init(struct ll_trunc_sem *sem)
301 atomic_set(&sem->ll_trunc_readers, 0);
302 atomic_set(&sem->ll_trunc_waiters, 0);
305 /* This version of down read ignores waiting writers, meaning if the semaphore
306 * is already held for read, this down_read will 'join' that reader and also
307 * take the semaphore.
309 * This lets us avoid an unusual deadlock.
311 * We must take lli_trunc_sem in read mode on entry in to various i/o paths
312 * in Lustre, in order to exclude truncates. Some of these paths then need to
313 * take the mmap_lock, while still holding the trunc_sem. The problem is that
314 * page faults hold the mmap_lock when calling in to Lustre, and then must also
315 * take the trunc_sem to exclude truncate.
317 * This means the locking order for trunc_sem and mmap_lock is sometimes AB,
318 * sometimes BA. This is almost OK because in both cases, we take the trunc
319 * sem for read, so it doesn't block.
321 * However, if a write mode user (truncate, a setattr op) arrives in the
322 * middle of this, the second reader on the truncate_sem will wait behind that
325 * So we have, on our truncate sem, in order (where 'reader' and 'writer' refer
326 * to the mode in which they take the semaphore):
327 * reader (holding mmap_lock, needs truncate_sem)
329 * reader (holding truncate sem, waiting for mmap_lock)
331 * And so the readers deadlock.
333 * The solution is this modified semaphore, where this down_read ignores
334 * waiting write operations, and all waiters are woken up at once, so readers
335 * using down_read_nowait cannot get stuck behind waiting writers, regardless
336 * of the order they arrived in.
338 * down_read_nowait is only used in the page fault case, where we already hold
339 * the mmap_lock. This is because otherwise repeated read and write operations
340 * (which take the truncate sem) could prevent a truncate from ever starting.
341 * This could still happen with page faults, but without an even more complex
342 * mechanism, this is unavoidable.
346 static inline void trunc_sem_down_read_nowait(struct ll_trunc_sem *sem)
348 wait_var_event(&sem->ll_trunc_readers,
349 atomic_inc_unless_negative(&sem->ll_trunc_readers));
352 static inline void trunc_sem_down_read(struct ll_trunc_sem *sem)
354 wait_var_event(&sem->ll_trunc_readers,
355 atomic_read(&sem->ll_trunc_waiters) == 0 &&
356 atomic_inc_unless_negative(&sem->ll_trunc_readers));
359 static inline void trunc_sem_up_read(struct ll_trunc_sem *sem)
361 if (atomic_dec_return(&sem->ll_trunc_readers) == 0 &&
362 atomic_read(&sem->ll_trunc_waiters))
363 wake_up_var(&sem->ll_trunc_readers);
366 static inline void trunc_sem_down_write(struct ll_trunc_sem *sem)
368 atomic_inc(&sem->ll_trunc_waiters);
369 wait_var_event(&sem->ll_trunc_readers,
370 atomic_cmpxchg(&sem->ll_trunc_readers, 0, -1) == 0);
371 atomic_dec(&sem->ll_trunc_waiters);
374 static inline void trunc_sem_up_write(struct ll_trunc_sem *sem)
376 atomic_set(&sem->ll_trunc_readers, 0);
377 /* match the smp_mb() in wait_var_event()->prepare_to_wait() */
379 wake_up_var(&sem->ll_trunc_readers);
382 #ifdef CONFIG_LUSTRE_FS_POSIX_ACL
383 static inline void lli_clear_acl(struct ll_inode_info *lli)
385 if (lli->lli_posix_acl) {
386 posix_acl_release(lli->lli_posix_acl);
387 lli->lli_posix_acl = NULL;
391 static inline void lli_replace_acl(struct ll_inode_info *lli,
392 struct lustre_md *md)
394 write_lock(&lli->lli_lock);
395 if (lli->lli_posix_acl)
396 posix_acl_release(lli->lli_posix_acl);
397 lli->lli_posix_acl = md->posix_acl;
398 write_unlock(&lli->lli_lock);
401 static inline void lli_clear_acl(struct ll_inode_info *lli)
405 static inline void lli_replace_acl(struct ll_inode_info *lli,
406 struct lustre_md *md)
411 static inline __u32 ll_layout_version_get(struct ll_inode_info *lli)
415 spin_lock(&lli->lli_layout_lock);
416 gen = lli->lli_layout_gen;
417 spin_unlock(&lli->lli_layout_lock);
422 static inline void ll_layout_version_set(struct ll_inode_info *lli, __u32 gen)
424 spin_lock(&lli->lli_layout_lock);
425 lli->lli_layout_gen = gen;
426 spin_unlock(&lli->lli_layout_lock);
430 /* File data is modified. */
431 LLIF_DATA_MODIFIED = 0,
432 /* File is being restored */
433 LLIF_FILE_RESTORING = 1,
434 /* Xattr cache is attached to the file */
435 LLIF_XATTR_CACHE = 2,
436 /* Project inherit */
437 LLIF_PROJECT_INHERIT = 3,
438 /* update atime from MDS even if it's older than local inode atime. */
439 LLIF_UPDATE_ATIME = 4,
440 /* foreign file/dir can be unlinked unconditionnaly */
441 LLIF_FOREIGN_REMOVABLE = 5,
442 /* Xattr cache is filled */
443 LLIF_XATTR_CACHE_FILLED = 7,
447 int ll_xattr_cache_destroy(struct inode *inode);
448 int ll_xattr_cache_empty(struct inode *inode);
450 int ll_xattr_cache_get(struct inode *inode,
456 int ll_xattr_cache_insert(struct inode *inode,
461 static inline bool obd_connect_has_secctx(struct obd_connect_data *data)
463 #ifdef CONFIG_SECURITY
464 return data->ocd_connect_flags & OBD_CONNECT_FLAGS2 &&
465 data->ocd_connect_flags2 & OBD_CONNECT2_FILE_SECCTX;
471 static inline void obd_connect_set_secctx(struct obd_connect_data *data)
473 #ifdef CONFIG_SECURITY
474 data->ocd_connect_flags2 |= OBD_CONNECT2_FILE_SECCTX;
478 /* Only smack and selinux is known to use security contexts */
479 static inline bool ll_xattr_is_seclabel(const char *name)
481 return !strcmp(name, XATTR_NAME_SELINUX) ||
482 !strcmp(name, XATTR_NAME_SMACK);
485 static inline bool ll_xattr_suffix_is_seclabel(const char *suffix)
487 return !strcmp(suffix, XATTR_SELINUX_SUFFIX) ||
488 !strcmp(suffix, XATTR_SMACK_SUFFIX);
491 int ll_dentry_init_security(struct dentry *dentry, int mode, struct qstr *name,
492 const char **secctx_name, __u32 *secctx_name_size,
493 void **secctx, __u32 *secctx_size);
494 int ll_inode_init_security(struct dentry *dentry, struct inode *inode,
497 int ll_inode_notifysecctx(struct inode *inode,
498 void *secctx, __u32 secctxlen);
500 void ll_secctx_name_free(struct ll_sb_info *sbi);
502 int ll_secctx_name_store(struct inode *in);
504 __u32 ll_secctx_name_get(struct ll_sb_info *sbi, const char **secctx_name);
506 int ll_security_secctx_name_filter(struct ll_sb_info *sbi, int xattr_type,
509 static inline bool obd_connect_has_enc(struct obd_connect_data *data)
511 #ifdef HAVE_LUSTRE_CRYPTO
512 return data->ocd_connect_flags & OBD_CONNECT_FLAGS2 &&
513 data->ocd_connect_flags2 & OBD_CONNECT2_ENCRYPT;
519 static inline void obd_connect_set_enc(struct obd_connect_data *data)
521 #ifdef HAVE_LUSTRE_CRYPTO
522 data->ocd_connect_flags2 |= OBD_CONNECT2_ENCRYPT;
526 static inline bool obd_connect_has_name_enc(struct obd_connect_data *data)
528 #ifdef HAVE_LUSTRE_CRYPTO
529 return data->ocd_connect_flags & OBD_CONNECT_FLAGS2 &&
530 data->ocd_connect_flags2 & OBD_CONNECT2_ENCRYPT_NAME;
536 static inline void obd_connect_set_name_enc(struct obd_connect_data *data)
538 #ifdef HAVE_LUSTRE_CRYPTO
539 data->ocd_connect_flags2 |= OBD_CONNECT2_ENCRYPT_NAME;
543 static inline bool obd_connect_has_enc_fid2path(struct obd_connect_data *data)
545 #ifdef HAVE_LUSTRE_CRYPTO
546 return data->ocd_connect_flags & OBD_CONNECT_FLAGS2 &&
547 data->ocd_connect_flags2 & OBD_CONNECT2_ENCRYPT_FID2PATH;
553 static inline void obd_connect_set_enc_fid2path(struct obd_connect_data *data)
555 #ifdef HAVE_LUSTRE_CRYPTO
556 data->ocd_connect_flags2 |= OBD_CONNECT2_ENCRYPT_FID2PATH;
561 * Locking to guarantee consistency of non-atomic updates to long long i_size,
562 * consistency between file size and KMS.
564 * Implemented by ->lli_size_mutex and ->lsm_lock, nested in that order.
567 void ll_inode_size_lock(struct inode *inode);
568 void ll_inode_size_unlock(struct inode *inode);
569 int ll_inode_size_trylock(struct inode *inode);
571 static inline struct ll_inode_info *ll_i2info(struct inode *inode)
573 return container_of(inode, struct ll_inode_info, lli_vfs_inode);
576 static inline struct pcc_inode *ll_i2pcci(struct inode *inode)
578 return ll_i2info(inode)->lli_pcc_inode;
581 /* default to use at least 16M for fast read if possible */
582 #define RA_REMAIN_WINDOW_MIN MiB_TO_PAGES(16UL)
584 /* default read-ahead on a given client mountpoint. */
585 #define SBI_DEFAULT_READ_AHEAD_MAX MiB_TO_PAGES(1024UL)
587 /* default read-ahead for a single file descriptor */
588 #define SBI_DEFAULT_READ_AHEAD_PER_FILE_MAX MiB_TO_PAGES(256UL)
590 /* default read-ahead full files smaller than limit on the second read */
591 #define SBI_DEFAULT_READ_AHEAD_WHOLE_MAX MiB_TO_PAGES(2UL)
593 /* default range pages */
594 #define SBI_DEFAULT_RA_RANGE_PAGES MiB_TO_PAGES(1ULL)
596 /* Min range pages */
597 #define RA_MIN_MMAP_RANGE_PAGES 16UL
602 RA_STAT_DISTANT_READPAGE,
603 RA_STAT_MISS_IN_WINDOW,
604 RA_STAT_FAILED_GRAB_PAGE,
605 RA_STAT_FAILED_MATCH,
610 RA_STAT_MAX_IN_FLIGHT,
611 RA_STAT_WRONG_GRAB_PAGE,
612 RA_STAT_FAILED_REACH_END,
614 RA_STAT_FAILED_FAST_READ,
615 RA_STAT_MMAP_RANGE_READ,
616 RA_STAT_READAHEAD_PAGES,
621 atomic_t ra_cur_pages;
622 unsigned long ra_max_pages;
623 unsigned long ra_max_pages_per_file;
624 unsigned long ra_range_pages;
625 unsigned long ra_max_read_ahead_whole_pages;
626 struct workqueue_struct *ll_readahead_wq;
628 * Max number of active works could be triggered
629 * for async readahead.
631 unsigned int ra_async_max_active;
632 /* how many async readahead triggered in flight */
633 atomic_t ra_async_inflight;
634 /* Threshold to control when to trigger async readahead */
635 unsigned long ra_async_pages_per_file_threshold;
638 /* ra_io_arg will be filled in the beginning of ll_readahead with
639 * ras_lock, then the following ll_read_ahead_pages will read RA
640 * pages according to this arg, all the items in this structure are
641 * counted by page index.
644 pgoff_t ria_start_idx; /* start offset of read-ahead*/
645 pgoff_t ria_end_idx; /* end offset of read-ahead*/
646 unsigned long ria_reserved; /* reserved pages for read-ahead */
647 pgoff_t ria_end_idx_min;/* minimum end to cover current read */
648 bool ria_eof; /* reach end of file */
649 /* If stride read pattern is detected, ria_stoff is the byte offset
650 * where stride read is started. Note: for normal read-ahead, the
651 * value here is meaningless, and also it will not be accessed*/
653 /* ria_length and ria_bytes are the length and pages length in the
654 * stride I/O mode. And they will also be used to check whether
655 * it is stride I/O read-ahead in the read-ahead pages*/
660 /* LL_HIST_MAX=32 causes an overflow */
661 #define LL_HIST_MAX 28
662 #define LL_HIST_START 12 /* buckets start at 2^12 = 4k */
663 #define LL_PROCESS_HIST_MAX 10
664 struct per_process_info {
666 struct obd_histogram pp_r_hist;
667 struct obd_histogram pp_w_hist;
670 /* pp_extents[LL_PROCESS_HIST_MAX] will hold the combined process info */
671 struct ll_rw_extents_info {
673 struct per_process_info pp_extents[LL_PROCESS_HIST_MAX + 1];
676 #define LL_OFFSET_HIST_MAX 100
677 struct ll_rw_process_info {
680 loff_t rw_range_start;
682 loff_t rw_last_file_pos;
684 size_t rw_smallest_extent;
685 size_t rw_largest_extent;
686 struct ll_file_data *rw_last_file;
689 enum stats_track_type {
690 STATS_TRACK_ALL = 0, /* track all processes */
691 STATS_TRACK_PID, /* track process with this pid */
692 STATS_TRACK_PPID, /* track processes with this ppid */
693 STATS_TRACK_GID, /* track processes with this gid */
697 /* flags for sbi->ll_flags */
699 LL_SBI_NOLCK, /* DLM locking disabled directio-only */
700 LL_SBI_CHECKSUM, /* checksum each page as it's written */
701 LL_SBI_LOCALFLOCK, /* local flocks instead of fs-wide */
702 LL_SBI_FLOCK, /* flock enabled */
703 LL_SBI_USER_XATTR, /* support user xattr */
704 LL_SBI_LRU_RESIZE, /* lru resize support */
705 LL_SBI_LAZYSTATFS, /* lazystatfs mount option */
706 LL_SBI_32BIT_API, /* generate 32 bit inodes. */
707 LL_SBI_USER_FID2PATH, /* fid2path by unprivileged users */
708 LL_SBI_VERBOSE, /* verbose mount/umount */
709 LL_SBI_ALWAYS_PING, /* ping even if server suppress_pings */
710 LL_SBI_TEST_DUMMY_ENCRYPTION, /* test dummy encryption */
711 LL_SBI_ENCRYPT, /* client side encryption */
712 LL_SBI_FOREIGN_SYMLINK, /* foreign fake-symlink support */
713 LL_SBI_FOREIGN_SYMLINK_UPCALL, /* foreign fake-symlink upcall set */
714 LL_SBI_NUM_MOUNT_OPT,
716 LL_SBI_ACL, /* support ACL */
717 LL_SBI_AGL_ENABLED, /* enable agl */
718 LL_SBI_64BIT_HASH, /* support 64-bits dir hash/offset */
719 LL_SBI_LAYOUT_LOCK, /* layout lock support */
720 LL_SBI_XATTR_CACHE, /* support for xattr cache */
721 LL_SBI_NOROOTSQUASH, /* do not apply root squash */
722 LL_SBI_FAST_READ, /* fast read support */
723 LL_SBI_FILE_SECCTX, /* file security context at create */
724 LL_SBI_TINY_WRITE, /* tiny write support */
725 LL_SBI_FILE_HEAT, /* file heat support */
726 LL_SBI_PARALLEL_DIO, /* parallel (async) O_DIRECT RPCs */
727 LL_SBI_ENCRYPT_NAME, /* name encryption */
731 int ll_sbi_flags_seq_show(struct seq_file *m, void *v);
733 /* This is embedded into llite super-blocks to keep track of connect
734 * flags (capabilities) supported by all imports given mount is
736 struct lustre_client_ocd {
737 /* This is conjunction of connect_flags across all imports
738 * (LOVs) this mount is connected to. This field is updated by
739 * cl_ocd_update() under ->lco_lock. */
741 struct mutex lco_lock;
742 struct obd_export *lco_md_exp;
743 struct obd_export *lco_dt_exp;
747 /* this protects pglist and ra_info. It isn't safe to
748 * grab from interrupt contexts */
750 spinlock_t ll_pp_extent_lock; /* pp_extent entry*/
751 spinlock_t ll_process_lock; /* ll_rw_process_info */
752 struct lustre_sb_info *lsi;
753 struct obd_uuid ll_sb_uuid;
754 struct obd_export *ll_md_exp;
755 struct obd_export *ll_dt_exp;
756 struct obd_device *ll_md_obd;
757 struct obd_device *ll_dt_obd;
758 struct dentry *ll_debugfs_entry;
759 struct lu_fid ll_root_fid; /* root object fid */
760 struct mnt_namespace *ll_mnt_ns;
762 DECLARE_BITMAP(ll_flags, LL_SBI_NUM_FLAGS); /* enum ll_sbi_flags */
763 unsigned int ll_xattr_cache_enabled:1,
764 ll_xattr_cache_set:1, /* already set to 0/1 */
765 ll_client_common_fill_super_succeeded:1,
767 ll_inode_cache_enabled:1;
769 struct lustre_client_ocd ll_lco;
771 struct lprocfs_stats *ll_stats; /* lprocfs stats counter */
773 /* Used to track "unstable" pages on a client, and maintain a
774 * LRU list of clean pages. An "unstable" page is defined as
775 * any page which is sent to a server as part of a bulk request,
776 * but is uncommitted to stable storage. */
777 struct cl_client_cache *ll_cache;
779 struct lprocfs_stats *ll_ra_stats;
781 struct ll_ra_info ll_ra_info;
782 unsigned int ll_namelen;
783 const struct file_operations *ll_fop;
785 struct lu_site *ll_site;
786 struct cl_device *ll_cl;
789 struct ll_rw_extents_info *ll_rw_extents_info;
790 int ll_extent_process_count;
791 unsigned int ll_offset_process_count;
792 struct ll_rw_process_info *ll_rw_process_info;
793 struct ll_rw_process_info *ll_rw_offset_info;
794 ktime_t ll_process_stats_init;
795 unsigned int ll_rw_offset_entry_count;
796 int ll_stats_track_id;
797 enum stats_track_type ll_stats_track_type;
800 /* metadata stat-ahead */
801 unsigned int ll_sa_running_max;/* max concurrent
802 * statahead instances */
803 unsigned int ll_sa_batch_max;/* max SUB request count in
804 * a batch PTLRPC request */
805 unsigned int ll_sa_max; /* max statahead RPCs */
806 atomic_t ll_sa_total; /* statahead thread started
808 atomic_t ll_sa_wrong; /* statahead thread stopped for
810 atomic_t ll_sa_running; /* running statahead thread
812 atomic_t ll_agl_total; /* AGL thread started count */
813 atomic_t ll_sa_hit_total; /* total hit count */
814 atomic_t ll_sa_miss_total; /* total miss count */
816 dev_t ll_sdev_orig; /* save s_dev before assign for
819 struct root_squash_info ll_squash;
822 /* st_blksize returned by stat(2), when non-zero */
823 unsigned int ll_stat_blksize;
825 /* maximum relative age of cached statfs results */
826 unsigned int ll_statfs_max_age;
828 struct kset ll_kset; /* sysfs object */
829 struct completion ll_kobj_unregister;
832 unsigned int ll_heat_decay_weight;
833 unsigned int ll_heat_period_second;
835 /* Opens of the same inode before we start requesting open lock */
836 u32 ll_oc_thrsh_count;
838 /* Time in ms between last inode close and next open to be considered
839 * instant back to back and would trigger an open lock request
843 /* Time in ms after last file close that we no longer count prior opens*/
846 /* filesystem fsname */
847 char ll_fsname[LUSTRE_MAXFSNAME + 1];
849 /* Persistent Client Cache */
850 struct pcc_super ll_pcc_super;
852 /* to protect vs updates in all following foreign symlink fields */
853 struct rw_semaphore ll_foreign_symlink_sem;
854 /* foreign symlink path prefix */
855 char *ll_foreign_symlink_prefix;
856 /* full prefix size including leading '\0' */
857 size_t ll_foreign_symlink_prefix_size;
858 /* foreign symlink path upcall */
859 char *ll_foreign_symlink_upcall;
860 /* foreign symlink path upcall infos */
861 struct ll_foreign_symlink_upcall_item *ll_foreign_symlink_upcall_items;
862 /* foreign symlink path upcall nb infos */
863 unsigned int ll_foreign_symlink_upcall_nb_items;
865 /* cached file security context xattr name. e.g: security.selinux */
866 char *ll_secctx_name;
867 __u32 ll_secctx_name_size;
870 #define SBI_DEFAULT_HEAT_DECAY_WEIGHT ((80 * 256 + 50) / 100)
871 #define SBI_DEFAULT_HEAT_PERIOD_SECOND (60)
873 #define SBI_DEFAULT_OPENCACHE_THRESHOLD_COUNT (5)
874 #define SBI_DEFAULT_OPENCACHE_THRESHOLD_MS (100) /* 0.1 second */
875 #define SBI_DEFAULT_OPENCACHE_THRESHOLD_MAX_MS (60000) /* 1 minute */
878 * per file-descriptor read-ahead data.
880 struct ll_readahead_state {
882 /* End byte that read(2) try to read. */
883 loff_t ras_last_read_end_bytes;
885 * number of bytes read after last read-ahead window reset. As window
886 * is reset on each seek, this is effectively a number of consecutive
887 * accesses. Maybe ->ras_accessed_in_window is better name.
889 * XXX nikita: window is also reset (by ras_update()) when Lustre
890 * believes that memory pressure evicts read-ahead pages. In that
891 * case, it probably doesn't make sense to expand window to
892 * PTLRPC_MAX_BRW_PAGES on the third access.
894 loff_t ras_consecutive_bytes;
896 * number of read requests after the last read-ahead window reset
897 * As window is reset on each seek, this is effectively the number
898 * on consecutive read request and is used to trigger read-ahead.
900 unsigned long ras_consecutive_requests;
902 * Parameters of current read-ahead window. Handled by
903 * ras_update(). On the initial access to the file or after a seek,
904 * window is reset to 0. After 3 consecutive accesses, window is
905 * expanded to PTLRPC_MAX_BRW_PAGES. Afterwards, window is enlarged by
906 * PTLRPC_MAX_BRW_PAGES chunks up to ->ra_max_pages.
908 pgoff_t ras_window_start_idx;
909 pgoff_t ras_window_pages;
911 /* Page index where min range read starts */
912 pgoff_t ras_range_min_start_idx;
913 /* Page index where mmap range read ends */
914 pgoff_t ras_range_max_end_idx;
915 /* number of mmap pages where last time detected */
916 pgoff_t ras_last_range_pages;
917 /* number of mmap range requests */
918 pgoff_t ras_range_requests;
921 * Optimal RPC size in pages.
922 * It decides how many pages will be sent for each read-ahead.
924 unsigned long ras_rpc_pages;
926 * Where next read-ahead should start at. This lies within read-ahead
927 * window. Read-ahead window is read in pieces rather than at once
928 * because: 1. lustre limits total number of pages under read-ahead by
929 * ->ra_max_pages (see ll_ra_count_get()), 2. client cannot read pages
930 * not covered by DLM lock.
932 pgoff_t ras_next_readahead_idx;
934 * Total number of ll_file_read requests issued, reads originating
935 * due to mmap are not counted in this total. This value is used to
936 * trigger full file read-ahead after multiple reads to a small file.
938 unsigned long ras_requests;
940 * The following 3 items are used for detecting the stride I/O
942 * In stride I/O mode,
943 * ...............|-----data-----|****gap*****|--------|******|....
944 * offset |-stride_bytes-|-stride_gap-|
945 * ras_stride_offset = offset;
946 * ras_stride_length = stride_bytes + stride_gap;
947 * ras_stride_bytes = stride_bytes;
948 * Note: all these three items are counted by bytes.
950 loff_t ras_stride_offset;
951 loff_t ras_stride_length;
952 loff_t ras_stride_bytes;
954 * number of consecutive stride request count, and it is similar as
955 * ras_consecutive_requests, but used for stride I/O mode.
956 * Note: only more than 2 consecutive stride request are detected,
957 * stride read-ahead will be enable
959 unsigned long ras_consecutive_stride_requests;
960 /* index of the last page that async readahead starts */
961 pgoff_t ras_async_last_readpage_idx;
962 /* whether we should increase readahead window */
963 bool ras_need_increase_window;
964 /* whether ra miss check should be skipped */
965 bool ras_no_miss_check;
968 struct ll_readahead_work {
969 /** File to readahead */
970 struct file *lrw_file;
971 pgoff_t lrw_start_idx;
975 /* async worker to handler read */
976 struct work_struct lrw_readahead_work;
977 char lrw_jobid[LUSTRE_JOBID_SIZE];
980 extern struct kmem_cache *ll_file_data_slab;
981 struct lustre_handle;
982 struct ll_file_data {
983 struct ll_readahead_state fd_ras;
984 struct ll_grouplock fd_grouplock;
988 /* openhandle if lease exists for this file.
989 * Borrow lli->lli_och_mutex to protect assignment */
990 struct obd_client_handle *fd_lease_och;
991 struct obd_client_handle *fd_och;
992 struct file *fd_file;
993 /* Indicate whether need to report failure when close.
994 * true: failure is known, not report again.
995 * false: unknown failure, should report. */
996 bool fd_write_failed;
997 bool ll_lock_no_expand;
998 /* Used by mirrored file to lead IOs to a specific mirror, usually
999 * for mirror resync. 0 means default. */
1000 __u32 fd_designated_mirror;
1001 /* The layout version when resync starts. Resync I/O should carry this
1002 * layout version for verification to OST objects */
1003 __u32 fd_layout_version;
1004 struct pcc_file fd_pcc_file;
1005 /* striped directory may read partially if some stripe inaccessible,
1006 * -errno is saved here, and will return to user in close().
1008 int fd_partial_readdir_rc;
1011 void llite_tunables_unregister(void);
1012 int llite_tunables_register(void);
1014 static inline struct inode *ll_info2i(struct ll_inode_info *lli)
1016 return &lli->lli_vfs_inode;
1019 __u32 ll_i2suppgid(struct inode *i);
1020 void ll_i2gids(__u32 *suppgids, struct inode *i1,struct inode *i2);
1022 static inline int ll_need_32bit_api(struct ll_sb_info *sbi)
1024 #if BITS_PER_LONG == 32
1026 #elif defined(CONFIG_COMPAT)
1027 if (unlikely(test_bit(LL_SBI_32BIT_API, sbi->ll_flags)))
1030 # ifdef CONFIG_X86_X32
1031 /* in_compat_syscall() returns true when called from a kthread
1032 * and CONFIG_X86_X32 is enabled, which is wrong. So check
1033 * whether the caller comes from a syscall (ie. not a kthread)
1034 * before calling in_compat_syscall(). */
1035 if (current->flags & PF_KTHREAD)
1039 return unlikely(in_compat_syscall());
1041 return unlikely(test_bit(LL_SBI_32BIT_API, sbi->ll_flags));
1045 static inline bool ll_sbi_has_fast_read(struct ll_sb_info *sbi)
1047 return test_bit(LL_SBI_FAST_READ, sbi->ll_flags);
1050 static inline bool ll_sbi_has_tiny_write(struct ll_sb_info *sbi)
1052 return test_bit(LL_SBI_TINY_WRITE, sbi->ll_flags);
1055 static inline bool ll_sbi_has_file_heat(struct ll_sb_info *sbi)
1057 return test_bit(LL_SBI_FILE_HEAT, sbi->ll_flags);
1060 static inline bool ll_sbi_has_foreign_symlink(struct ll_sb_info *sbi)
1062 return test_bit(LL_SBI_FOREIGN_SYMLINK, sbi->ll_flags);
1065 static inline bool ll_sbi_has_parallel_dio(struct ll_sb_info *sbi)
1067 return test_bit(LL_SBI_PARALLEL_DIO, sbi->ll_flags);
1070 void ll_ras_enter(struct file *f, loff_t pos, size_t count);
1072 /* llite/lcommon_misc.c */
1073 int cl_ocd_update(struct obd_device *host, struct obd_device *watched,
1074 enum obd_notify_event ev, void *owner);
1075 int cl_get_grouplock(struct cl_object *obj, unsigned long gid, int nonblock,
1076 struct ll_grouplock *lg);
1077 void cl_put_grouplock(struct ll_grouplock *lg);
1079 /* llite/lproc_llite.c */
1080 int ll_debugfs_register_super(struct super_block *sb, const char *name);
1081 void ll_debugfs_unregister_super(struct super_block *sb);
1082 void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, long count);
1083 void ll_free_rw_stats_info(struct ll_sb_info *sbi);
1086 LPROC_LL_READ_BYTES,
1087 LPROC_LL_WRITE_BYTES,
1114 LPROC_LL_GETXATTR_HITS,
1116 LPROC_LL_REMOVEXATTR,
1117 LPROC_LL_INODE_PERM,
1119 LPROC_LL_INODE_OCOUNT,
1120 LPROC_LL_INODE_OPCLTM,
1121 LPROC_LL_FILE_OPCODES
1125 enum get_default_layout_type {
1126 GET_DEFAULT_LAYOUT_ROOT = 1,
1129 extern const struct file_operations ll_dir_operations;
1130 extern const struct inode_operations ll_dir_inode_operations;
1131 #ifdef HAVE_DIR_CONTEXT
1132 int ll_dir_read(struct inode *inode, __u64 *pos, struct md_op_data *op_data,
1133 struct dir_context *ctx, int *partial_readdir_rc);
1135 int ll_dir_read(struct inode *inode, __u64 *pos, struct md_op_data *op_data,
1136 void *cookie, filldir_t filldir, int *partial_readdir_rc);
1138 int ll_get_mdt_idx(struct inode *inode);
1139 int ll_get_mdt_idx_by_fid(struct ll_sb_info *sbi, const struct lu_fid *fid);
1140 struct page *ll_get_dir_page(struct inode *dir, struct md_op_data *op_data,
1141 __u64 offset, int *partial_readdir_rc);
1142 void ll_release_page(struct inode *inode, struct page *page, bool remove);
1143 int quotactl_ioctl(struct super_block *sb, struct if_quotactl *qctl);
1146 extern const struct inode_operations ll_special_inode_operations;
1148 struct inode *ll_iget(struct super_block *sb, ino_t hash,
1149 struct lustre_md *lic);
1150 int ll_test_inode_by_fid(struct inode *inode, void *opaque);
1151 int ll_md_blocking_ast(struct ldlm_lock *, struct ldlm_lock_desc *,
1152 void *data, int flag);
1153 struct dentry *ll_splice_alias(struct inode *inode, struct dentry *de);
1154 int ll_rmdir_entry(struct inode *dir, char *name, int namelen);
1155 void ll_update_times(struct ptlrpc_request *request, struct inode *inode);
1158 int ll_writepage(struct page *page, struct writeback_control *wbc);
1159 int ll_writepages(struct address_space *, struct writeback_control *wbc);
1160 int ll_readpage(struct file *file, struct page *page);
1161 #ifdef HAVE_AOPS_READ_FOLIO
1162 int ll_read_folio(struct file *file, struct folio *folio);
1164 int ll_io_read_page(const struct lu_env *env, struct cl_io *io,
1165 struct cl_page *page, struct file *file);
1166 void ll_readahead_init(struct inode *inode, struct ll_readahead_state *ras);
1167 int vvp_io_write_commit(const struct lu_env *env, struct cl_io *io);
1170 void ll_cl_add(struct inode *inode, const struct lu_env *env, struct cl_io *io,
1171 enum lcc_type type);
1172 void ll_cl_remove(struct inode *inode, const struct lu_env *env);
1173 struct ll_cl_context *ll_cl_find(struct inode *inode);
1175 extern const struct address_space_operations ll_aops;
1178 extern const struct inode_operations ll_file_inode_operations;
1179 const struct file_operations *ll_select_file_operations(struct ll_sb_info *sbi);
1180 extern int ll_have_md_lock(struct obd_export *exp, struct inode *inode,
1181 __u64 *bits, enum ldlm_mode l_req_mode);
1182 extern enum ldlm_mode ll_take_md_lock(struct inode *inode, __u64 bits,
1183 struct lustre_handle *lockh, __u64 flags,
1184 enum ldlm_mode mode);
1186 int ll_file_open(struct inode *inode, struct file *file);
1187 int ll_file_release(struct inode *inode, struct file *file);
1188 int ll_release_openhandle(struct dentry *, struct lookup_intent *);
1189 int ll_md_real_close(struct inode *inode, fmode_t fmode);
1190 void ll_track_file_opens(struct inode *inode);
1191 extern void ll_rw_stats_tally(struct ll_sb_info *sbi, pid_t pid,
1192 struct ll_file_data *file, loff_t pos,
1193 size_t count, int rw);
1194 #if defined(HAVE_USER_NAMESPACE_ARG) || defined(HAVE_INODEOPS_ENHANCED_GETATTR)
1195 int ll_getattr(struct user_namespace *mnt_userns, const struct path *path,
1196 struct kstat *stat, u32 request_mask, unsigned int flags);
1198 int ll_getattr(struct vfsmount *mnt, struct dentry *de, struct kstat *stat);
1199 #endif /* HAVE_USER_NAMESPACE_ARG */
1200 int ll_getattr_dentry(struct dentry *de, struct kstat *stat, u32 request_mask,
1201 unsigned int flags, bool foreign);
1202 #ifdef CONFIG_LUSTRE_FS_POSIX_ACL
1203 struct posix_acl *ll_get_acl(struct inode *inode, int type
1204 #ifdef HAVE_GET_ACL_RCU_ARG
1206 #endif /* HAVE_GET_ACL_RCU_ARG */
1208 int ll_set_acl(struct user_namespace *mnt_userns, struct inode *inode,
1209 struct posix_acl *acl, int type);
1210 #else /* !CONFIG_LUSTRE_FS_POSIX_ACL */
1211 #define ll_get_acl NULL
1212 #define ll_set_acl NULL
1213 #endif /* CONFIG_LUSTRE_FS_POSIX_ACL */
1215 static inline int ll_xflags_to_inode_flags(int xflags)
1217 return ((xflags & FS_XFLAG_SYNC) ? S_SYNC : 0) |
1218 ((xflags & FS_XFLAG_NOATIME) ? S_NOATIME : 0) |
1219 ((xflags & FS_XFLAG_APPEND) ? S_APPEND : 0) |
1220 ((xflags & FS_XFLAG_IMMUTABLE) ? S_IMMUTABLE : 0);
1223 static inline int ll_inode_flags_to_xflags(int inode_flags)
1225 return ((inode_flags & S_SYNC) ? FS_XFLAG_SYNC : 0) |
1226 ((inode_flags & S_NOATIME) ? FS_XFLAG_NOATIME : 0) |
1227 ((inode_flags & S_APPEND) ? FS_XFLAG_APPEND : 0) |
1228 ((inode_flags & S_IMMUTABLE) ? FS_XFLAG_IMMUTABLE : 0);
1231 int ll_migrate(struct inode *parent, struct file *file,
1232 struct lmv_user_md *lum, const char *name, __u32 flags);
1233 int ll_get_fid_by_name(struct inode *parent, const char *name,
1234 int namelen, struct lu_fid *fid, struct inode **inode);
1235 int ll_inode_permission(struct user_namespace *mnt_userns, struct inode *inode,
1237 int ll_ioctl_check_project(struct inode *inode, __u32 xflags, __u32 projid);
1238 int ll_ioctl_fsgetxattr(struct inode *inode, unsigned int cmd,
1240 int ll_ioctl_fssetxattr(struct inode *inode, unsigned int cmd,
1242 int ll_ioctl_project(struct file *file, unsigned int cmd, void __user *uarg);
1243 int ll_lov_setstripe_ea_info(struct inode *inode, struct dentry *dentry,
1244 __u64 flags, struct lov_user_md *lum,
1246 int ll_lov_getstripe_ea_info(struct inode *inode, const char *filename,
1247 struct lov_mds_md **lmm, int *lmm_size,
1248 struct ptlrpc_request **request);
1249 int ll_dir_setstripe(struct inode *inode, struct lov_user_md *lump,
1251 int ll_dir_get_default_layout(struct inode *inode, void **plmm, int *plmm_size,
1252 struct ptlrpc_request **request, u64 valid,
1253 enum get_default_layout_type type);
1254 int ll_dir_getstripe_default(struct inode *inode, void **lmmp,
1255 int *lmm_size, struct ptlrpc_request **request,
1256 struct ptlrpc_request **root_request, u64 valid);
1257 int ll_dir_getstripe(struct inode *inode, void **plmm, int *plmm_size,
1258 struct ptlrpc_request **request, u64 valid);
1259 int ll_fsync(struct file *file, loff_t start, loff_t end, int data);
1260 int ll_merge_attr(const struct lu_env *env, struct inode *inode);
1261 int ll_merge_attr_try(const struct lu_env *env, struct inode *inode);
1262 int ll_fid2path(struct inode *inode, void __user *arg);
1263 int __ll_fid2path(struct inode *inode, struct getinfo_fid2path *gfout,
1264 size_t outsize, __u32 pathlen_orig);
1265 int ll_data_version(struct inode *inode, __u64 *data_version, int flags);
1266 int ll_hsm_release(struct inode *inode);
1267 int ll_hsm_state_set(struct inode *inode, struct hsm_state_set *hss);
1268 void ll_io_set_mirror(struct cl_io *io, const struct file *file);
1270 /* llite/dcache.c */
1272 extern const struct dentry_operations ll_d_ops;
1274 bool ll_d_setup(struct dentry *de, bool do_put);
1276 static inline bool lld_is_init(struct dentry *dentry)
1278 return ll_d2d(dentry);
1281 #define ll_d_setup(de, do_put) (true)
1282 #define lld_is_init(dentry) (true)
1285 void ll_intent_drop_lock(struct lookup_intent *);
1286 void ll_intent_release(struct lookup_intent *);
1287 void ll_prune_aliases(struct inode *inode);
1288 void ll_lookup_finish_locks(struct lookup_intent *it, struct dentry *dentry);
1289 int ll_revalidate_it_finish(struct ptlrpc_request *request,
1290 struct lookup_intent *it, struct dentry *de);
1292 /* llite/llite_lib.c */
1293 extern const struct super_operations lustre_super_operations;
1295 void ll_lli_init(struct ll_inode_info *lli);
1296 int ll_fill_super(struct super_block *sb);
1297 void ll_put_super(struct super_block *sb);
1298 void ll_kill_super(struct super_block *sb);
1299 struct inode *ll_inode_from_resource_lock(struct ldlm_lock *lock);
1300 void ll_dir_clear_lsm_md(struct inode *inode);
1301 void ll_clear_inode(struct inode *inode);
1302 int volatile_ref_file(const char *volatile_name, int volatile_len,
1303 struct file **ref_file);
1304 int ll_setattr_raw(struct dentry *dentry, struct iattr *attr,
1305 enum op_xvalid xvalid, bool hsm_import);
1306 int ll_setattr(struct user_namespace *mnt_userns, struct dentry *de,
1307 struct iattr *attr);
1308 int ll_statfs(struct dentry *de, struct kstatfs *sfs);
1309 int ll_statfs_internal(struct ll_sb_info *sbi, struct obd_statfs *osfs,
1311 int ll_update_inode(struct inode *inode, struct lustre_md *md);
1312 void ll_update_inode_flags(struct inode *inode, unsigned int ext_flags);
1313 void ll_update_dir_depth(struct inode *dir, struct inode *inode);
1314 int ll_read_inode2(struct inode *inode, void *opaque);
1315 void ll_truncate_inode_pages_final(struct inode *inode, struct cl_io *io);
1316 void ll_delete_inode(struct inode *inode);
1317 int ll_iocontrol(struct inode *inode, struct file *file,
1318 unsigned int cmd, void __user *uarg);
1319 int ll_flush_ctx(struct inode *inode);
1320 void ll_umount_begin(struct super_block *sb);
1321 int ll_remount_fs(struct super_block *sb, int *flags, char *data);
1322 int ll_show_options(struct seq_file *seq, struct dentry *dentry);
1323 void ll_dirty_page_discard_warn(struct inode *inode, int ioret);
1324 int ll_prep_inode(struct inode **inode, struct req_capsule *pill,
1325 struct super_block *sb, struct lookup_intent *it);
1326 int ll_obd_statfs(struct inode *inode, void __user *arg);
1327 int ll_get_max_mdsize(struct ll_sb_info *sbi, int *max_mdsize);
1328 int ll_get_default_mdsize(struct ll_sb_info *sbi, int *default_mdsize);
1329 int ll_set_default_mdsize(struct ll_sb_info *sbi, int default_mdsize);
1331 void ll_unlock_md_op_lsm(struct md_op_data *op_data);
1332 struct md_op_data *ll_prep_md_op_data(struct md_op_data *op_data,
1333 struct inode *i1, struct inode *i2,
1334 const char *name, size_t namelen,
1335 __u32 mode, enum md_op_code opc,
1337 void ll_finish_md_op_data(struct md_op_data *op_data);
1338 int ll_get_obd_name(struct inode *inode, unsigned int cmd, void __user *arg);
1339 void ll_compute_rootsquash_state(struct ll_sb_info *sbi);
1340 ssize_t ll_copy_user_md(const struct lov_user_md __user *md,
1341 struct lov_user_md **kbuf);
1342 void ll_open_cleanup(struct super_block *sb, struct req_capsule *pill);
1344 void ll_dom_finish_open(struct inode *inode, struct ptlrpc_request *req);
1346 /* Compute expected user md size when passing in a md from user space */
1347 static inline ssize_t ll_lov_user_md_size(const struct lov_user_md *lum)
1349 switch (lum->lmm_magic) {
1350 case LOV_USER_MAGIC_V1:
1351 return sizeof(struct lov_user_md_v1);
1352 case LOV_USER_MAGIC_V3:
1353 return sizeof(struct lov_user_md_v3);
1354 case LOV_USER_MAGIC_SPECIFIC:
1355 if (lum->lmm_stripe_count > LOV_MAX_STRIPE_COUNT)
1358 return lov_user_md_size(lum->lmm_stripe_count,
1359 LOV_USER_MAGIC_SPECIFIC);
1360 case LOV_USER_MAGIC_COMP_V1:
1361 return ((struct lov_comp_md_v1 *)lum)->lcm_size;
1362 case LOV_USER_MAGIC_FOREIGN:
1363 return foreign_size(lum);
1369 /* llite/llite_nfs.c */
1370 extern const struct export_operations lustre_export_operations;
1371 __u32 get_uuid2int(const char *name, int len);
1372 struct inode *search_inode_for_lustre(struct super_block *sb,
1373 const struct lu_fid *fid);
1374 int ll_dir_get_parent_fid(struct inode *dir, struct lu_fid *parent_fid);
1376 /* llite/symlink.c */
1377 extern const struct inode_operations ll_fast_symlink_inode_operations;
1380 * IO arguments for various VFS I/O interfaces.
1382 struct vvp_io_args {
1383 /** normal/sendfile/splice */
1386 struct kiocb *via_iocb;
1387 struct iov_iter *via_iter;
1397 struct ll_cl_context {
1398 struct list_head lcc_list;
1400 const struct lu_env *lcc_env;
1401 struct cl_io *lcc_io;
1402 struct cl_page *lcc_page;
1403 enum lcc_type lcc_type;
1404 pgoff_t lcc_end_index;
1407 struct ll_thread_info {
1408 struct vvp_io_args lti_args;
1409 struct ra_io_arg lti_ria;
1410 struct ll_cl_context lti_io_ctx;
1413 extern struct lu_context_key ll_thread_key;
1415 static inline struct ll_thread_info *ll_env_info(const struct lu_env *env)
1417 struct ll_thread_info *lti;
1419 lti = lu_context_key_get(&env->le_ctx, &ll_thread_key);
1420 LASSERT(lti != NULL);
1425 static inline struct vvp_io_args *ll_env_args(const struct lu_env *env)
1427 return &ll_env_info(env)->lti_args;
1430 void ll_io_init(struct cl_io *io, struct file *file, enum cl_io_type iot,
1431 struct vvp_io_args *args);
1433 /* llite/llite_mmap.c */
1435 int ll_file_mmap(struct file * file, struct vm_area_struct * vma);
1436 void policy_from_vma(union ldlm_policy_data *policy, struct vm_area_struct *vma,
1437 unsigned long addr, size_t count);
1438 struct vm_area_struct *our_vma(struct mm_struct *mm, unsigned long addr,
1441 #define ll_s2sbi(sb) (s2lsi(sb)->lsi_llsbi)
1443 /* don't need an addref as the sb_info should be holding one */
1444 static inline struct obd_export *ll_s2dtexp(struct super_block *sb)
1446 return ll_s2sbi(sb)->ll_dt_exp;
1449 /* don't need an addref as the sb_info should be holding one */
1450 static inline struct obd_export *ll_s2mdexp(struct super_block *sb)
1452 return ll_s2sbi(sb)->ll_md_exp;
1455 static inline struct client_obd *sbi2mdc(struct ll_sb_info *sbi)
1457 struct obd_device *obd = sbi->ll_md_exp->exp_obd;
1463 // FIXME: replace the name of this with LL_SB to conform to kernel stuff
1464 static inline struct ll_sb_info *ll_i2sbi(struct inode *inode)
1466 return ll_s2sbi(inode->i_sb);
1469 static inline struct obd_export *ll_i2dtexp(struct inode *inode)
1471 return ll_s2dtexp(inode->i_sb);
1474 static inline struct obd_export *ll_i2mdexp(struct inode *inode)
1476 return ll_s2mdexp(inode->i_sb);
1479 static inline struct lu_fid *ll_inode2fid(struct inode *inode)
1483 LASSERT(inode != NULL);
1484 fid = &ll_i2info(inode)->lli_fid;
1489 static inline bool ll_dir_striped(struct inode *inode)
1491 struct ll_inode_info *lli;
1495 if (!S_ISDIR(inode->i_mode))
1498 lli = ll_i2info(inode);
1499 if (!lli->lli_lsm_md)
1502 down_read(&lli->lli_lsm_sem);
1503 rc = lmv_dir_striped(lli->lli_lsm_md);
1504 up_read(&lli->lli_lsm_sem);
1509 static inline loff_t ll_file_maxbytes(struct inode *inode)
1511 struct cl_object *obj = ll_i2info(inode)->lli_clob;
1514 return MAX_LFS_FILESIZE;
1516 return min_t(loff_t, cl_object_maxbytes(obj), MAX_LFS_FILESIZE);
1520 extern const struct xattr_handler *ll_xattr_handlers[];
1522 #define XATTR_USER_T 1
1523 #define XATTR_TRUSTED_T 2
1524 #define XATTR_SECURITY_T 3
1525 #define XATTR_ACL_ACCESS_T 4
1526 #define XATTR_ACL_DEFAULT_T 5
1527 #define XATTR_LUSTRE_T 6
1528 #define XATTR_OTHER_T 7
1529 #define XATTR_ENCRYPTION_T 9
1531 ssize_t ll_listxattr(struct dentry *dentry, char *buffer, size_t size);
1532 int ll_xattr_list(struct inode *inode, const char *name, int type,
1533 void *buffer, size_t size, u64 valid);
1534 const struct xattr_handler *get_xattr_type(const char *name);
1537 * Common IO arguments for various VFS I/O interfaces.
1539 int cl_sb_init(struct super_block *sb);
1540 int cl_sb_fini(struct super_block *sb);
1542 enum ras_update_flags {
1546 void ll_ra_count_put(struct ll_sb_info *sbi, unsigned long len);
1547 void ll_ra_stats_inc(struct inode *inode, enum ra_stat which);
1551 #define LL_SA_RPC_MIN 8
1552 #define LL_SA_RPC_DEF 32
1553 #define LL_SA_RPC_MAX 2048
1555 /* XXX: If want to support more concurrent statahead instances,
1556 * please consider to decentralize the RPC lists attached
1557 * on related import, such as imp_{sending,delayed}_list.
1559 #define LL_SA_RUNNING_MAX 256
1560 #define LL_SA_RUNNING_DEF 16
1562 #define LL_SA_BATCH_MAX 1024
1563 #define LL_SA_BATCH_DEF 0
1565 #define LL_SA_CACHE_BIT 5
1566 #define LL_SA_CACHE_SIZE (1 << LL_SA_CACHE_BIT)
1567 #define LL_SA_CACHE_MASK (LL_SA_CACHE_SIZE - 1)
1569 /* per inode struct, for dir only */
1570 struct ll_statahead_info {
1571 struct dentry *sai_dentry;
1572 atomic_t sai_refcount; /* when access this struct, hold
1574 unsigned int sai_max; /* max ahead of lookup */
1575 __u64 sai_sent; /* stat requests sent count */
1576 __u64 sai_replied; /* stat requests which received
1578 __u64 sai_index; /* index of statahead entry */
1579 __u64 sai_index_wait; /* index of entry which is the
1580 * caller is waiting for */
1581 __u64 sai_hit; /* hit count */
1582 __u64 sai_miss; /* miss count:
1583 * for "ls -al" case, includes
1584 * hidden dentry miss;
1585 * for "ls -l" case, it does not
1586 * include hidden dentry miss.
1587 * "sai_miss_hidden" is used for
1590 unsigned int sai_consecutive_miss; /* consecutive miss */
1591 unsigned int sai_miss_hidden;/* "ls -al", but first dentry
1592 * is not a hidden one */
1593 unsigned int sai_skip_hidden;/* skipped hidden dentry count
1595 unsigned int sai_ls_all:1; /* "ls -al", do stat-ahead for
1597 wait_queue_head_t sai_waitq; /* stat-ahead wait queue */
1598 struct task_struct *sai_task; /* stat-ahead thread */
1599 struct task_struct *sai_agl_task; /* AGL thread */
1600 struct list_head sai_entries; /* completed entries */
1601 struct list_head sai_agls; /* AGLs to be sent */
1602 struct list_head sai_cache[LL_SA_CACHE_SIZE];
1603 spinlock_t sai_cache_lock[LL_SA_CACHE_SIZE];
1604 atomic_t sai_cache_count; /* entry count in cache */
1605 struct lu_batch *sai_bh;
1606 __u32 sai_max_batch_count;
1607 __u64 sai_index_end;
1610 int ll_revalidate_statahead(struct inode *dir, struct dentry **dentry,
1612 int ll_start_statahead(struct inode *dir, struct dentry *dentry, bool agl);
1613 void ll_authorize_statahead(struct inode *dir, void *key);
1614 void ll_deauthorize_statahead(struct inode *dir, void *key);
1617 blkcnt_t dirty_cnt(struct inode *inode);
1619 int cl_glimpse_size0(struct inode *inode, int agl);
1620 int cl_glimpse_lock(const struct lu_env *env, struct cl_io *io,
1621 struct inode *inode, struct cl_object *clob, int agl);
1623 static inline int cl_glimpse_size(struct inode *inode)
1625 return cl_glimpse_size0(inode, 0);
1628 /* AGL is 'asychronous glimpse lock', which is a speculative lock taken as
1629 * part of statahead */
1630 static inline int cl_agl(struct inode *inode)
1632 return cl_glimpse_size0(inode, 1);
1635 int ll_file_lock_ahead(struct file *file, struct llapi_lu_ladvise *ladvise);
1637 int cl_io_get(struct inode *inode, struct lu_env **envout,
1638 struct cl_io **ioout, __u16 *refcheck);
1640 static inline int ll_glimpse_size(struct inode *inode)
1642 struct ll_inode_info *lli = ll_i2info(inode);
1645 down_read(&lli->lli_glimpse_sem);
1646 rc = cl_glimpse_size(inode);
1647 lli->lli_glimpse_time = ktime_get();
1648 up_read(&lli->lli_glimpse_sem);
1652 /* dentry may statahead when statahead is enabled and current process has opened
1653 * parent directory, and this dentry hasn't accessed statahead cache before */
1655 dentry_may_statahead(struct inode *dir, struct dentry *dentry)
1657 struct ll_inode_info *lli;
1658 struct ll_dentry_data *ldd;
1660 if (ll_i2sbi(dir)->ll_sa_max == 0)
1663 lli = ll_i2info(dir);
1665 /* statahead is not allowed for this dir, there may be three causes:
1666 * 1. dir is not opened.
1667 * 2. statahead hit ratio is too low.
1668 * 3. previous stat started statahead thread failed. */
1669 if (!lli->lli_sa_enabled)
1672 /* not the same process, don't statahead */
1673 if (lli->lli_opendir_pid != current->pid)
1677 * When stating a dentry, kernel may trigger 'revalidate' or 'lookup'
1678 * multiple times, eg. for 'getattr', 'getxattr' and etc.
1679 * For patchless client, lookup intent is not accurate, which may
1680 * misguide statahead. For example:
1681 * The 'revalidate' call for 'getattr' and 'getxattr' of a dentry will
1682 * have the same intent -- IT_GETATTR, while one dentry should access
1683 * statahead cache once, otherwise statahead windows is messed up.
1684 * The solution is as following:
1685 * Assign 'lld_sa_generation' with 'lli_sa_generation' when a dentry
1686 * IT_GETATTR for the first time, and subsequent IT_GETATTR will
1687 * bypass interacting with statahead cache by checking
1688 * 'lld_sa_generation == lli->lli_sa_generation'.
1690 ldd = ll_d2d(dentry);
1691 if (ldd != NULL && lli->lli_sa_generation &&
1692 ldd->lld_sa_generation == lli->lli_sa_generation)
1698 int cl_sync_file_range(struct inode *inode, loff_t start, loff_t end,
1699 enum cl_fsync_mode mode, int ignore_layout);
1701 static inline int ll_file_nolock(const struct file *file)
1703 struct ll_file_data *fd = file->private_data;
1704 struct inode *inode = file_inode((struct file *)file);
1706 LASSERT(fd != NULL);
1707 return ((fd->fd_flags & LL_FILE_IGNORE_LOCK) ||
1708 test_bit(LL_SBI_NOLCK, ll_i2sbi(inode)->ll_flags));
1711 static inline void ll_set_lock_data(struct obd_export *exp, struct inode *inode,
1712 struct lookup_intent *it, __u64 *bits)
1714 if (!it->it_lock_set) {
1715 struct lustre_handle handle;
1717 /* If this inode is a remote object, it will get two
1718 * separate locks in different namespaces, Master MDT,
1719 * where the name entry is, will grant LOOKUP lock,
1720 * remote MDT, where the object is, will grant
1721 * UPDATE|PERM lock. The inode will be attched to both
1722 * LOOKUP and PERM locks, so revoking either locks will
1723 * case the dcache being cleared */
1724 if (it->it_remote_lock_mode) {
1725 handle.cookie = it->it_remote_lock_handle;
1726 CDEBUG(D_DLMTRACE, "setting l_data to inode "DFID
1727 "(%p) for remote lock %#llx\n",
1728 PFID(ll_inode2fid(inode)), inode,
1730 md_set_lock_data(exp, &handle, inode, NULL);
1733 handle.cookie = it->it_lock_handle;
1735 CDEBUG(D_DLMTRACE, "setting l_data to inode "DFID"(%p)"
1736 " for lock %#llx\n",
1737 PFID(ll_inode2fid(inode)), inode, handle.cookie);
1739 md_set_lock_data(exp, &handle, inode, &it->it_lock_bits);
1740 it->it_lock_set = 1;
1744 *bits = it->it_lock_bits;
1747 static inline int d_lustre_invalid(const struct dentry *dentry)
1749 return !ll_d2d(dentry) || ll_d2d(dentry)->lld_invalid;
1753 * Mark dentry INVALID, if dentry refcount is zero (this is normally case for
1754 * ll_md_blocking_ast), it will be pruned by ll_prune_aliases() and
1755 * ll_prune_negative_children(); otherwise dput() of the last refcount will
1756 * unhash this dentry and kill it.
1758 static inline void d_lustre_invalidate(struct dentry *dentry)
1760 CDEBUG(D_DENTRY, "invalidate dentry %pd (%p) parent %p inode %p refc %d\n",
1762 dentry->d_parent, dentry->d_inode, ll_d_count(dentry));
1764 spin_lock(&dentry->d_lock);
1765 if (lld_is_init(dentry))
1766 ll_d2d(dentry)->lld_invalid = 1;
1767 spin_unlock(&dentry->d_lock);
1770 static inline void d_lustre_revalidate(struct dentry *dentry)
1772 spin_lock(&dentry->d_lock);
1773 LASSERT(ll_d2d(dentry));
1774 ll_d2d(dentry)->lld_invalid = 0;
1775 spin_unlock(&dentry->d_lock);
1778 static inline dev_t ll_compat_encode_dev(dev_t dev)
1780 /* The compat_sys_*stat*() syscalls will fail unless the
1781 * device majors and minors are both less than 256. Note that
1782 * the value returned here will be passed through
1783 * old_encode_dev() in cp_compat_stat(). And so we are not
1784 * trying to return a valid compat (u16) device number, just
1785 * one that will pass the old_valid_dev() check. */
1787 return MKDEV(MAJOR(dev) & 0xff, MINOR(dev) & 0xff);
1790 int ll_layout_conf(struct inode *inode, const struct cl_object_conf *conf);
1791 int ll_layout_refresh(struct inode *inode, __u32 *gen);
1792 int ll_layout_restore(struct inode *inode, loff_t start, __u64 length);
1793 int ll_layout_write_intent(struct inode *inode, enum layout_intent_opc opc,
1794 struct lu_extent *ext);
1796 int ll_xattr_init(void);
1797 void ll_xattr_fini(void);
1799 int ll_page_sync_io(const struct lu_env *env, struct cl_io *io,
1800 struct cl_page *page, enum cl_req_type crt);
1802 int ll_getparent(struct file *file, struct getparent __user *arg);
1805 int cl_setattr_ost(struct cl_object *obj, const struct iattr *attr,
1806 enum op_xvalid xvalid, unsigned int attr_flags);
1808 extern struct lu_env *cl_inode_fini_env;
1809 extern __u16 cl_inode_fini_refcheck;
1811 int cl_file_inode_init(struct inode *inode, struct lustre_md *md);
1812 void cl_inode_fini(struct inode *inode);
1814 u64 cl_fid_build_ino(const struct lu_fid *fid, int api32);
1815 u32 cl_fid_build_gen(const struct lu_fid *fid);
1817 static inline struct pcc_super *ll_i2pccs(struct inode *inode)
1819 return &ll_i2sbi(inode)->ll_pcc_super;
1822 static inline struct pcc_super *ll_info2pccs(struct ll_inode_info *lli)
1824 return ll_i2pccs(ll_info2i(lli));
1828 /* The digested form is made of a FID (16 bytes) followed by the second-to-last
1829 * ciphertext block (16 bytes), so a total length of 32 bytes.
1830 * That way, llcrypt does not compute a digested form of this digest.
1832 struct ll_digest_filename {
1833 struct lu_fid ldf_fid;
1834 char ldf_excerpt[LL_CRYPTO_BLOCK_SIZE];
1837 int ll_prepare_lookup(struct inode *dir, struct dentry *de,
1838 struct llcrypt_name *fname, struct lu_fid *fid);
1839 int ll_setup_filename(struct inode *dir, const struct qstr *iname,
1840 int lookup, struct llcrypt_name *fname,
1841 struct lu_fid *fid);
1842 int ll_fname_disk_to_usr(struct inode *inode,
1843 u32 hash, u32 minor_hash,
1844 struct llcrypt_str *iname, struct llcrypt_str *oname,
1845 struct lu_fid *fid);
1846 int ll_file_open_encrypt(struct inode *inode, struct file *filp);
1848 static inline char *xattr_for_enc(struct inode *inode)
1850 if (ll_sbi_has_name_encrypt(ll_i2sbi(inode)))
1851 return LL_XATTR_NAME_ENCRYPTION_CONTEXT;
1853 return LL_XATTR_NAME_ENCRYPTION_CONTEXT_OLD;
1856 #ifdef HAVE_LUSTRE_CRYPTO
1857 extern const struct llcrypt_operations lustre_cryptops;
1860 /* llite/llite_foreign.c */
1861 int ll_manage_foreign(struct inode *inode, struct lustre_md *lmd);
1862 bool ll_foreign_is_openable(struct dentry *dentry, unsigned int flags);
1863 bool ll_foreign_is_removable(struct dentry *dentry, bool unset);
1865 int ll_filemap_fault(struct vm_area_struct *vma, struct vm_fault *vmf);
1867 #endif /* LLITE_INTERNAL_H */