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 /* default LMV is explicitly set in inode on MDT, this
192 * is for old server, or default LMV is set by
193 * "lfs setdirstripe -D".
195 lli_default_lmv_set:1;
196 /* generation for statahead */
197 unsigned int lli_sa_generation;
198 /* rw lock protects lli_lsm_md */
199 struct rw_semaphore lli_lsm_sem;
200 /* directory stripe information */
201 struct lmv_stripe_md *lli_lsm_md;
202 /* directory default LMV */
203 struct lmv_stripe_md *lli_default_lsm_md;
206 /* for non-directory */
208 struct mutex lli_size_mutex;
209 struct task_struct *lli_size_lock_owner;
210 char *lli_symlink_name;
211 struct ll_trunc_sem lli_trunc_sem;
212 struct range_lock_tree lli_write_tree;
213 struct mutex lli_setattr_mutex;
215 struct rw_semaphore lli_glimpse_sem;
216 ktime_t lli_glimpse_time;
217 struct list_head lli_agl_list;
220 /* for writepage() only to communicate to fsync */
223 /* protect the file heat fields */
224 spinlock_t lli_heat_lock;
225 __u32 lli_heat_flags;
226 struct obd_heat_instance lli_heat_instances[OBD_HEAT_COUNT];
229 * Whenever a process try to read/write the file, the
230 * jobid, uid and gid of the process will be saved here,
231 * and it'll be packed into write RPC when flush later.
233 * So the read/write statistics or TBF rules for jobid,
234 * uid or gid will not be accurate if the file is shared
237 char lli_jobid[LUSTRE_JOBID_SIZE];
241 struct mutex lli_pcc_lock;
242 enum lu_pcc_state_flags lli_pcc_state;
244 * @lli_pcc_generation saves the gobal PCC generation
245 * when the file was successfully attached into PCC.
246 * The flags of the PCC dataset are saved in
248 * The gobal PCC generation will be increased when add
249 * or delete a PCC backend, or change the configuration
250 * parameters for PCC.
251 * If @lli_pcc_generation is same as the gobal PCC
252 * generation, we can use the saved flags of the PCC
253 * dataset to determine whether need to try auto attach
256 __u64 lli_pcc_generation;
257 enum pcc_dataset_flags lli_pcc_dsflags;
258 struct pcc_inode *lli_pcc_inode;
260 struct mutex lli_group_mutex;
261 __u64 lli_group_users;
262 unsigned long lli_group_gid;
264 __u64 lli_attr_valid;
266 __u64 lli_lazyblocks;
270 /* XXX: For following frequent used members, although they maybe special
271 * used for non-directory object, it is some time-wasting to check
272 * whether the object is directory or not before using them. On the
273 * other hand, currently, sizeof(f) > sizeof(d), it cannot reduce
274 * the "ll_inode_info" size even if moving those members into u.f.
275 * So keep them out side.
277 * In the future, if more members are added only for directory,
278 * some of the following members can be moved into u.f.
280 struct cl_object *lli_clob;
282 /* mutex to request for layout lock exclusively. */
283 struct mutex lli_layout_mutex;
284 /* Layout version, protected by lli_layout_lock */
285 __u32 lli_layout_gen;
286 spinlock_t lli_layout_lock;
288 __u32 lli_projid; /* project id */
290 struct rw_semaphore lli_xattrs_list_rwsem;
291 struct mutex lli_xattrs_enq_lock;
292 struct list_head lli_xattrs; /* ll_xattr_entry->xe_list */
293 struct list_head lli_lccs; /* list of ll_cl_context */
294 seqlock_t lli_page_inv_lock;
296 struct task_struct *lli_inode_lock_owner;
299 #ifndef HAVE_USER_NAMESPACE_ARG
300 #define inode_permission(ns, inode, mask) inode_permission(inode, mask)
301 #define generic_permission(ns, inode, mask) generic_permission(inode, mask)
302 #define simple_setattr(ns, de, iattr) simple_setattr(de, iattr)
303 #define ll_inode_permission(ns, inode, mask) ll_inode_permission(inode, mask)
304 #ifdef HAVE_INODEOPS_ENHANCED_GETATTR
305 #define ll_getattr(ns, path, stat, mask, fl) ll_getattr(path, stat, mask, fl)
306 #endif /* HAVE_INODEOPS_ENHANCED_GETATTR */
307 #define ll_setattr(ns, de, attr) ll_setattr(de, attr)
310 static inline void ll_trunc_sem_init(struct ll_trunc_sem *sem)
312 atomic_set(&sem->ll_trunc_readers, 0);
313 atomic_set(&sem->ll_trunc_waiters, 0);
316 /* This version of down read ignores waiting writers, meaning if the semaphore
317 * is already held for read, this down_read will 'join' that reader and also
318 * take the semaphore.
320 * This lets us avoid an unusual deadlock.
322 * We must take lli_trunc_sem in read mode on entry in to various i/o paths
323 * in Lustre, in order to exclude truncates. Some of these paths then need to
324 * take the mmap_lock, while still holding the trunc_sem. The problem is that
325 * page faults hold the mmap_lock when calling in to Lustre, and then must also
326 * take the trunc_sem to exclude truncate.
328 * This means the locking order for trunc_sem and mmap_lock is sometimes AB,
329 * sometimes BA. This is almost OK because in both cases, we take the trunc
330 * sem for read, so it doesn't block.
332 * However, if a write mode user (truncate, a setattr op) arrives in the
333 * middle of this, the second reader on the truncate_sem will wait behind that
336 * So we have, on our truncate sem, in order (where 'reader' and 'writer' refer
337 * to the mode in which they take the semaphore):
338 * reader (holding mmap_lock, needs truncate_sem)
340 * reader (holding truncate sem, waiting for mmap_lock)
342 * And so the readers deadlock.
344 * The solution is this modified semaphore, where this down_read ignores
345 * waiting write operations, and all waiters are woken up at once, so readers
346 * using down_read_nowait cannot get stuck behind waiting writers, regardless
347 * of the order they arrived in.
349 * down_read_nowait is only used in the page fault case, where we already hold
350 * the mmap_lock. This is because otherwise repeated read and write operations
351 * (which take the truncate sem) could prevent a truncate from ever starting.
352 * This could still happen with page faults, but without an even more complex
353 * mechanism, this is unavoidable.
357 static inline void trunc_sem_down_read_nowait(struct ll_trunc_sem *sem)
359 wait_var_event(&sem->ll_trunc_readers,
360 atomic_inc_unless_negative(&sem->ll_trunc_readers));
363 static inline void trunc_sem_down_read(struct ll_trunc_sem *sem)
365 wait_var_event(&sem->ll_trunc_readers,
366 atomic_read(&sem->ll_trunc_waiters) == 0 &&
367 atomic_inc_unless_negative(&sem->ll_trunc_readers));
370 static inline void trunc_sem_up_read(struct ll_trunc_sem *sem)
372 if (atomic_dec_return(&sem->ll_trunc_readers) == 0 &&
373 atomic_read(&sem->ll_trunc_waiters))
374 wake_up_var(&sem->ll_trunc_readers);
377 static inline void trunc_sem_down_write(struct ll_trunc_sem *sem)
379 atomic_inc(&sem->ll_trunc_waiters);
380 wait_var_event(&sem->ll_trunc_readers,
381 atomic_cmpxchg(&sem->ll_trunc_readers, 0, -1) == 0);
382 atomic_dec(&sem->ll_trunc_waiters);
385 static inline void trunc_sem_up_write(struct ll_trunc_sem *sem)
387 atomic_set(&sem->ll_trunc_readers, 0);
388 /* match the smp_mb() in wait_var_event()->prepare_to_wait() */
390 wake_up_var(&sem->ll_trunc_readers);
393 #ifdef CONFIG_LUSTRE_FS_POSIX_ACL
394 static inline void lli_clear_acl(struct ll_inode_info *lli)
396 if (lli->lli_posix_acl) {
397 posix_acl_release(lli->lli_posix_acl);
398 lli->lli_posix_acl = NULL;
402 static inline void lli_replace_acl(struct ll_inode_info *lli,
403 struct lustre_md *md)
405 write_lock(&lli->lli_lock);
406 if (lli->lli_posix_acl)
407 posix_acl_release(lli->lli_posix_acl);
408 lli->lli_posix_acl = md->posix_acl;
409 write_unlock(&lli->lli_lock);
412 static inline void lli_clear_acl(struct ll_inode_info *lli)
416 static inline void lli_replace_acl(struct ll_inode_info *lli,
417 struct lustre_md *md)
422 static inline __u32 ll_layout_version_get(struct ll_inode_info *lli)
426 spin_lock(&lli->lli_layout_lock);
427 gen = lli->lli_layout_gen;
428 spin_unlock(&lli->lli_layout_lock);
433 static inline void ll_layout_version_set(struct ll_inode_info *lli, __u32 gen)
435 spin_lock(&lli->lli_layout_lock);
436 lli->lli_layout_gen = gen;
437 spin_unlock(&lli->lli_layout_lock);
440 enum ll_inode_flags {
441 /* File data is modified. */
442 LLIF_DATA_MODIFIED = 0,
443 /* File is being restored */
444 LLIF_FILE_RESTORING = 1,
445 /* Xattr cache is attached to the file */
446 LLIF_XATTR_CACHE = 2,
447 /* Project inherit */
448 LLIF_PROJECT_INHERIT = 3,
449 /* update atime from MDS even if it's older than local inode atime. */
450 LLIF_UPDATE_ATIME = 4,
451 /* foreign file/dir can be unlinked unconditionnaly */
452 LLIF_FOREIGN_REMOVABLE = 5,
453 /* 6 is not used for now */
454 /* Xattr cache is filled */
455 LLIF_XATTR_CACHE_FILLED = 7,
459 int ll_xattr_cache_destroy(struct inode *inode);
460 int ll_xattr_cache_empty(struct inode *inode);
462 int ll_xattr_cache_get(struct inode *inode,
468 int ll_xattr_cache_insert(struct inode *inode,
473 static inline bool obd_connect_has_secctx(struct obd_connect_data *data)
475 #ifdef CONFIG_SECURITY
476 return data->ocd_connect_flags & OBD_CONNECT_FLAGS2 &&
477 data->ocd_connect_flags2 & OBD_CONNECT2_FILE_SECCTX;
483 static inline void obd_connect_set_secctx(struct obd_connect_data *data)
485 #ifdef CONFIG_SECURITY
486 data->ocd_connect_flags2 |= OBD_CONNECT2_FILE_SECCTX;
490 /* Only smack and selinux is known to use security contexts */
491 static inline bool ll_xattr_is_seclabel(const char *name)
493 return !strcmp(name, XATTR_NAME_SELINUX) ||
494 !strcmp(name, XATTR_NAME_SMACK);
497 static inline bool ll_xattr_suffix_is_seclabel(const char *suffix)
499 return !strcmp(suffix, XATTR_SELINUX_SUFFIX) ||
500 !strcmp(suffix, XATTR_SMACK_SUFFIX);
503 int ll_dentry_init_security(struct dentry *dentry, int mode, struct qstr *name,
504 const char **secctx_name, __u32 *secctx_name_size,
505 void **secctx, __u32 *secctx_size,
508 int ll_inode_init_security(struct dentry *dentry, struct inode *inode,
511 int ll_inode_notifysecctx(struct inode *inode,
512 void *secctx, __u32 secctxlen);
514 void ll_secctx_name_free(struct ll_sb_info *sbi);
516 int ll_secctx_name_store(struct inode *in);
518 __u32 ll_secctx_name_get(struct ll_sb_info *sbi, const char **secctx_name);
520 int ll_security_secctx_name_filter(struct ll_sb_info *sbi, int xattr_type,
523 static inline bool obd_connect_has_enc(struct obd_connect_data *data)
525 #ifdef HAVE_LUSTRE_CRYPTO
526 return data->ocd_connect_flags & OBD_CONNECT_FLAGS2 &&
527 data->ocd_connect_flags2 & OBD_CONNECT2_ENCRYPT;
533 static inline void obd_connect_set_enc(struct obd_connect_data *data)
535 #ifdef HAVE_LUSTRE_CRYPTO
536 data->ocd_connect_flags2 |= OBD_CONNECT2_ENCRYPT;
540 static inline bool obd_connect_has_name_enc(struct obd_connect_data *data)
542 #ifdef HAVE_LUSTRE_CRYPTO
543 return data->ocd_connect_flags & OBD_CONNECT_FLAGS2 &&
544 data->ocd_connect_flags2 & OBD_CONNECT2_ENCRYPT_NAME;
550 static inline void obd_connect_set_name_enc(struct obd_connect_data *data)
552 #ifdef HAVE_LUSTRE_CRYPTO
553 data->ocd_connect_flags2 |= OBD_CONNECT2_ENCRYPT_NAME;
557 static inline bool obd_connect_has_enc_fid2path(struct obd_connect_data *data)
559 #ifdef HAVE_LUSTRE_CRYPTO
560 return data->ocd_connect_flags & OBD_CONNECT_FLAGS2 &&
561 data->ocd_connect_flags2 & OBD_CONNECT2_ENCRYPT_FID2PATH;
567 static inline void obd_connect_set_enc_fid2path(struct obd_connect_data *data)
569 #ifdef HAVE_LUSTRE_CRYPTO
570 data->ocd_connect_flags2 |= OBD_CONNECT2_ENCRYPT_FID2PATH;
575 * Locking to guarantee consistency of non-atomic updates to long long i_size,
576 * consistency between file size and KMS.
578 * Implemented by ->lli_size_mutex and ->lsm_lock, nested in that order.
581 void ll_inode_size_lock(struct inode *inode);
582 void ll_inode_size_unlock(struct inode *inode);
583 int ll_inode_size_trylock(struct inode *inode);
585 static inline struct ll_inode_info *ll_i2info(struct inode *inode)
587 return container_of(inode, struct ll_inode_info, lli_vfs_inode);
590 static inline struct pcc_inode *ll_i2pcci(struct inode *inode)
592 return ll_i2info(inode)->lli_pcc_inode;
595 static inline void ll_set_inode_lock_owner(struct inode *inode)
597 ll_i2info(inode)->lli_inode_lock_owner = current;
600 static inline void ll_clear_inode_lock_owner(struct inode *inode)
602 ll_i2info(inode)->lli_inode_lock_owner = NULL;
605 static inline struct task_struct *ll_get_inode_lock_owner(struct inode *inode)
607 return ll_i2info(inode)->lli_inode_lock_owner;
610 /* lock inode and set inode lock owener */
611 static inline void ll_inode_lock(struct inode *inode)
614 ll_set_inode_lock_owner(inode);
617 /* clear inode lock owner and unlock it */
618 static inline void ll_inode_unlock(struct inode *inode)
620 ll_clear_inode_lock_owner(inode);
624 /* default to use at least 16M for fast read if possible */
625 #define RA_REMAIN_WINDOW_MIN MiB_TO_PAGES(16UL)
627 /* default read-ahead on a given client mountpoint. */
628 #define SBI_DEFAULT_READ_AHEAD_MAX MiB_TO_PAGES(1024UL)
630 /* default read-ahead for a single file descriptor */
631 #define SBI_DEFAULT_READ_AHEAD_PER_FILE_MAX MiB_TO_PAGES(256UL)
633 /* default read-ahead full files smaller than limit on the second read */
634 #define SBI_DEFAULT_READ_AHEAD_WHOLE_MAX MiB_TO_PAGES(2UL)
636 /* default range pages */
637 #define SBI_DEFAULT_RA_RANGE_PAGES MiB_TO_PAGES(1ULL)
639 /* Min range pages */
640 #define RA_MIN_MMAP_RANGE_PAGES 16UL
645 RA_STAT_DISTANT_READPAGE,
646 RA_STAT_MISS_IN_WINDOW,
647 RA_STAT_FAILED_GRAB_PAGE,
648 RA_STAT_FAILED_MATCH,
653 RA_STAT_MAX_IN_FLIGHT,
654 RA_STAT_WRONG_GRAB_PAGE,
655 RA_STAT_FAILED_REACH_END,
657 RA_STAT_FAILED_FAST_READ,
658 RA_STAT_MMAP_RANGE_READ,
659 RA_STAT_READAHEAD_PAGES,
664 atomic_t ra_cur_pages;
665 unsigned long ra_max_pages;
666 unsigned long ra_max_pages_per_file;
667 unsigned long ra_range_pages;
668 unsigned long ra_max_read_ahead_whole_pages;
669 struct workqueue_struct *ll_readahead_wq;
671 * Max number of active works could be triggered
672 * for async readahead.
674 unsigned int ra_async_max_active;
675 /* how many async readahead triggered in flight */
676 atomic_t ra_async_inflight;
677 /* Threshold to control when to trigger async readahead */
678 unsigned long ra_async_pages_per_file_threshold;
681 /* ra_io_arg will be filled in the beginning of ll_readahead with
682 * ras_lock, then the following ll_read_ahead_pages will read RA
683 * pages according to this arg, all the items in this structure are
684 * counted by page index.
687 pgoff_t ria_start_idx; /* start offset of read-ahead*/
688 pgoff_t ria_end_idx; /* end offset of read-ahead*/
689 unsigned long ria_reserved; /* reserved pages for read-ahead */
690 pgoff_t ria_end_idx_min;/* minimum end to cover current read */
691 bool ria_eof; /* reach end of file */
692 /* If stride read pattern is detected, ria_stoff is the byte offset
693 * where stride read is started. Note: for normal read-ahead, the
694 * value here is meaningless, and also it will not be accessed*/
696 /* ria_length and ria_bytes are the length and pages length in the
697 * stride I/O mode. And they will also be used to check whether
698 * it is stride I/O read-ahead in the read-ahead pages*/
703 /* LL_HIST_MAX=32 causes an overflow */
704 #define LL_HIST_MAX 28
705 #define LL_HIST_START 12 /* buckets start at 2^12 = 4k */
706 #define LL_PROCESS_HIST_MAX 10
707 struct per_process_info {
709 struct obd_histogram pp_r_hist;
710 struct obd_histogram pp_w_hist;
713 /* pp_extents[LL_PROCESS_HIST_MAX] will hold the combined process info */
714 struct ll_rw_extents_info {
716 struct per_process_info pp_extents[LL_PROCESS_HIST_MAX + 1];
719 #define LL_OFFSET_HIST_MAX 100
720 struct ll_rw_process_info {
723 loff_t rw_range_start;
725 loff_t rw_last_file_pos;
727 size_t rw_smallest_extent;
728 size_t rw_largest_extent;
729 struct ll_file_data *rw_last_file;
732 enum stats_track_type {
733 STATS_TRACK_ALL = 0, /* track all processes */
734 STATS_TRACK_PID, /* track process with this pid */
735 STATS_TRACK_PPID, /* track processes with this ppid */
736 STATS_TRACK_GID, /* track processes with this gid */
740 /* flags for sbi->ll_flags */
742 LL_SBI_NOLCK, /* DLM locking disabled directio-only */
743 LL_SBI_CHECKSUM, /* checksum each page as it's written */
744 LL_SBI_LOCALFLOCK, /* local flocks instead of fs-wide */
745 LL_SBI_FLOCK, /* flock enabled */
746 LL_SBI_USER_XATTR, /* support user xattr */
747 LL_SBI_LRU_RESIZE, /* lru resize support */
748 LL_SBI_LAZYSTATFS, /* lazystatfs mount option */
749 LL_SBI_32BIT_API, /* generate 32 bit inodes. */
750 LL_SBI_USER_FID2PATH, /* fid2path by unprivileged users */
751 LL_SBI_VERBOSE, /* verbose mount/umount */
752 LL_SBI_ALWAYS_PING, /* ping even if server suppress_pings */
753 LL_SBI_TEST_DUMMY_ENCRYPTION, /* test dummy encryption */
754 LL_SBI_ENCRYPT, /* client side encryption */
755 LL_SBI_FOREIGN_SYMLINK, /* foreign fake-symlink support */
756 LL_SBI_FOREIGN_SYMLINK_UPCALL, /* foreign fake-symlink upcall set */
757 LL_SBI_NUM_MOUNT_OPT,
759 LL_SBI_ACL, /* support ACL */
760 LL_SBI_AGL_ENABLED, /* enable agl */
761 LL_SBI_64BIT_HASH, /* support 64-bits dir hash/offset */
762 LL_SBI_LAYOUT_LOCK, /* layout lock support */
763 LL_SBI_XATTR_CACHE, /* support for xattr cache */
764 LL_SBI_NOROOTSQUASH, /* do not apply root squash */
765 LL_SBI_FAST_READ, /* fast read support */
766 LL_SBI_FILE_SECCTX, /* file security context at create */
767 LL_SBI_TINY_WRITE, /* tiny write support */
768 LL_SBI_FILE_HEAT, /* file heat support */
769 LL_SBI_PARALLEL_DIO, /* parallel (async) O_DIRECT RPCs */
770 LL_SBI_ENCRYPT_NAME, /* name encryption */
774 int ll_sbi_flags_seq_show(struct seq_file *m, void *v);
776 /* This is embedded into llite super-blocks to keep track of connect
777 * flags (capabilities) supported by all imports given mount is
779 struct lustre_client_ocd {
780 /* This is conjunction of connect_flags across all imports
781 * (LOVs) this mount is connected to. This field is updated by
782 * cl_ocd_update() under ->lco_lock. */
784 struct mutex lco_lock;
785 struct obd_export *lco_md_exp;
786 struct obd_export *lco_dt_exp;
790 /* this protects pglist and ra_info. It isn't safe to
791 * grab from interrupt contexts */
793 spinlock_t ll_pp_extent_lock; /* pp_extent entry*/
794 spinlock_t ll_process_lock; /* ll_rw_process_info */
795 struct lustre_sb_info *lsi;
796 struct obd_uuid ll_sb_uuid;
797 struct obd_export *ll_md_exp;
798 struct obd_export *ll_dt_exp;
799 struct obd_device *ll_md_obd;
800 struct obd_device *ll_dt_obd;
801 struct dentry *ll_debugfs_entry;
802 struct lu_fid ll_root_fid; /* root object fid */
803 struct mnt_namespace *ll_mnt_ns;
805 DECLARE_BITMAP(ll_flags, LL_SBI_NUM_FLAGS); /* enum ll_sbi_flags */
806 unsigned int ll_xattr_cache_enabled:1,
807 ll_xattr_cache_set:1, /* already set to 0/1 */
808 ll_client_common_fill_super_succeeded:1,
810 ll_inode_cache_enabled:1;
812 struct lustre_client_ocd ll_lco;
814 struct lprocfs_stats *ll_stats; /* lprocfs stats counter */
816 /* Used to track "unstable" pages on a client, and maintain a
817 * LRU list of clean pages. An "unstable" page is defined as
818 * any page which is sent to a server as part of a bulk request,
819 * but is uncommitted to stable storage. */
820 struct cl_client_cache *ll_cache;
822 struct lprocfs_stats *ll_ra_stats;
824 struct ll_ra_info ll_ra_info;
825 unsigned int ll_namelen;
826 const struct file_operations *ll_fop;
828 struct lu_site *ll_site;
829 struct cl_device *ll_cl;
832 struct ll_rw_extents_info *ll_rw_extents_info;
833 int ll_extent_process_count;
834 unsigned int ll_offset_process_count;
835 struct ll_rw_process_info *ll_rw_process_info;
836 struct ll_rw_process_info *ll_rw_offset_info;
837 ktime_t ll_process_stats_init;
838 unsigned int ll_rw_offset_entry_count;
839 int ll_stats_track_id;
840 enum stats_track_type ll_stats_track_type;
843 /* metadata stat-ahead */
844 unsigned int ll_sa_running_max;/* max concurrent
845 * statahead instances */
846 unsigned int ll_sa_batch_max;/* max SUB request count in
847 * a batch PTLRPC request */
848 unsigned int ll_sa_max; /* max statahead RPCs */
849 atomic_t ll_sa_total; /* statahead thread started
851 atomic_t ll_sa_wrong; /* statahead thread stopped for
853 atomic_t ll_sa_running; /* running statahead thread
855 atomic_t ll_agl_total; /* AGL thread started count */
856 atomic_t ll_sa_hit_total; /* total hit count */
857 atomic_t ll_sa_miss_total; /* total miss count */
859 dev_t ll_sdev_orig; /* save s_dev before assign for
862 struct root_squash_info ll_squash;
865 /* st_blksize returned by stat(2), when non-zero */
866 unsigned int ll_stat_blksize;
868 /* maximum relative age of cached statfs results */
869 unsigned int ll_statfs_max_age;
871 struct kset ll_kset; /* sysfs object */
872 struct completion ll_kobj_unregister;
875 unsigned int ll_heat_decay_weight;
876 unsigned int ll_heat_period_second;
878 /* Opens of the same inode before we start requesting open lock */
879 u32 ll_oc_thrsh_count;
881 /* Time in ms between last inode close and next open to be considered
882 * instant back to back and would trigger an open lock request
886 /* Time in ms after last file close that we no longer count prior opens*/
889 /* filesystem fsname */
890 char ll_fsname[LUSTRE_MAXFSNAME + 1];
892 /* Persistent Client Cache */
893 struct pcc_super ll_pcc_super;
895 /* to protect vs updates in all following foreign symlink fields */
896 struct rw_semaphore ll_foreign_symlink_sem;
897 /* foreign symlink path prefix */
898 char *ll_foreign_symlink_prefix;
899 /* full prefix size including leading '\0' */
900 size_t ll_foreign_symlink_prefix_size;
901 /* foreign symlink path upcall */
902 char *ll_foreign_symlink_upcall;
903 /* foreign symlink path upcall infos */
904 struct ll_foreign_symlink_upcall_item *ll_foreign_symlink_upcall_items;
905 /* foreign symlink path upcall nb infos */
906 unsigned int ll_foreign_symlink_upcall_nb_items;
908 /* cached file security context xattr name. e.g: security.selinux */
909 char *ll_secctx_name;
910 __u32 ll_secctx_name_size;
913 #define SBI_DEFAULT_HEAT_DECAY_WEIGHT ((80 * 256 + 50) / 100)
914 #define SBI_DEFAULT_HEAT_PERIOD_SECOND (60)
916 #define SBI_DEFAULT_OPENCACHE_THRESHOLD_COUNT (5)
917 #define SBI_DEFAULT_OPENCACHE_THRESHOLD_MS (100) /* 0.1 second */
918 #define SBI_DEFAULT_OPENCACHE_THRESHOLD_MAX_MS (60000) /* 1 minute */
921 * per file-descriptor read-ahead data.
923 struct ll_readahead_state {
925 /* End byte that read(2) try to read. */
926 loff_t ras_last_read_end_bytes;
928 * number of bytes read after last read-ahead window reset. As window
929 * is reset on each seek, this is effectively a number of consecutive
930 * accesses. Maybe ->ras_accessed_in_window is better name.
932 * XXX nikita: window is also reset (by ras_update()) when Lustre
933 * believes that memory pressure evicts read-ahead pages. In that
934 * case, it probably doesn't make sense to expand window to
935 * PTLRPC_MAX_BRW_PAGES on the third access.
937 loff_t ras_consecutive_bytes;
939 * number of read requests after the last read-ahead window reset
940 * As window is reset on each seek, this is effectively the number
941 * on consecutive read request and is used to trigger read-ahead.
943 unsigned long ras_consecutive_requests;
945 * Parameters of current read-ahead window. Handled by
946 * ras_update(). On the initial access to the file or after a seek,
947 * window is reset to 0. After 3 consecutive accesses, window is
948 * expanded to PTLRPC_MAX_BRW_PAGES. Afterwards, window is enlarged by
949 * PTLRPC_MAX_BRW_PAGES chunks up to ->ra_max_pages.
951 pgoff_t ras_window_start_idx;
952 pgoff_t ras_window_pages;
954 /* Page index where min range read starts */
955 pgoff_t ras_range_min_start_idx;
956 /* Page index where mmap range read ends */
957 pgoff_t ras_range_max_end_idx;
958 /* number of mmap pages where last time detected */
959 pgoff_t ras_last_range_pages;
960 /* number of mmap range requests */
961 pgoff_t ras_range_requests;
964 * Optimal RPC size in pages.
965 * It decides how many pages will be sent for each read-ahead.
967 unsigned long ras_rpc_pages;
969 * Where next read-ahead should start at. This lies within read-ahead
970 * window. Read-ahead window is read in pieces rather than at once
971 * because: 1. lustre limits total number of pages under read-ahead by
972 * ->ra_max_pages (see ll_ra_count_get()), 2. client cannot read pages
973 * not covered by DLM lock.
975 pgoff_t ras_next_readahead_idx;
977 * Total number of ll_file_read requests issued, reads originating
978 * due to mmap are not counted in this total. This value is used to
979 * trigger full file read-ahead after multiple reads to a small file.
981 unsigned long ras_requests;
983 * The following 3 items are used for detecting the stride I/O
985 * In stride I/O mode,
986 * ...............|-----data-----|****gap*****|--------|******|....
987 * offset |-stride_bytes-|-stride_gap-|
988 * ras_stride_offset = offset;
989 * ras_stride_length = stride_bytes + stride_gap;
990 * ras_stride_bytes = stride_bytes;
991 * Note: all these three items are counted by bytes.
993 loff_t ras_stride_offset;
994 loff_t ras_stride_length;
995 loff_t ras_stride_bytes;
997 * number of consecutive stride request count, and it is similar as
998 * ras_consecutive_requests, but used for stride I/O mode.
999 * Note: only more than 2 consecutive stride request are detected,
1000 * stride read-ahead will be enable
1002 unsigned long ras_consecutive_stride_requests;
1003 /* index of the last page that async readahead starts */
1004 pgoff_t ras_async_last_readpage_idx;
1005 /* whether we should increase readahead window */
1006 bool ras_need_increase_window;
1007 /* whether ra miss check should be skipped */
1008 bool ras_no_miss_check;
1011 struct ll_readahead_work {
1012 /** File to readahead */
1013 struct file *lrw_file;
1014 pgoff_t lrw_start_idx;
1015 pgoff_t lrw_end_idx;
1018 /* async worker to handler read */
1019 struct work_struct lrw_readahead_work;
1020 char lrw_jobid[LUSTRE_JOBID_SIZE];
1023 extern struct kmem_cache *ll_file_data_slab;
1024 struct lustre_handle;
1025 struct ll_file_data {
1026 struct ll_readahead_state fd_ras;
1027 struct ll_grouplock fd_grouplock;
1031 /* openhandle if lease exists for this file.
1032 * Borrow lli->lli_och_mutex to protect assignment */
1033 struct obd_client_handle *fd_lease_och;
1034 struct obd_client_handle *fd_och;
1035 struct file *fd_file;
1036 /* Indicate whether need to report failure when close.
1037 * true: failure is known, not report again.
1038 * false: unknown failure, should report. */
1039 bool fd_write_failed;
1040 bool ll_lock_no_expand;
1041 /* Used by mirrored file to lead IOs to a specific mirror, usually
1042 * for mirror resync. 0 means default. */
1043 __u32 fd_designated_mirror;
1044 /* The layout version when resync starts. Resync I/O should carry this
1045 * layout version for verification to OST objects */
1046 __u32 fd_layout_version;
1047 struct pcc_file fd_pcc_file;
1048 /* striped directory may read partially if some stripe inaccessible,
1049 * -errno is saved here, and will return to user in close().
1051 int fd_partial_readdir_rc;
1054 void llite_tunables_unregister(void);
1055 int llite_tunables_register(void);
1057 static inline struct inode *ll_info2i(struct ll_inode_info *lli)
1059 return &lli->lli_vfs_inode;
1062 __u32 ll_i2suppgid(struct inode *i);
1063 void ll_i2gids(__u32 *suppgids, struct inode *i1,struct inode *i2);
1065 static inline int ll_need_32bit_api(struct ll_sb_info *sbi)
1067 #if BITS_PER_LONG == 32
1069 #elif defined(CONFIG_COMPAT)
1070 if (unlikely(test_bit(LL_SBI_32BIT_API, sbi->ll_flags)))
1073 # ifdef CONFIG_X86_X32
1074 /* in_compat_syscall() returns true when called from a kthread
1075 * and CONFIG_X86_X32 is enabled, which is wrong. So check
1076 * whether the caller comes from a syscall (ie. not a kthread)
1077 * before calling in_compat_syscall(). */
1078 if (current->flags & PF_KTHREAD)
1082 return unlikely(in_compat_syscall());
1084 return unlikely(test_bit(LL_SBI_32BIT_API, sbi->ll_flags));
1088 static inline bool ll_sbi_has_fast_read(struct ll_sb_info *sbi)
1090 return test_bit(LL_SBI_FAST_READ, sbi->ll_flags);
1093 static inline bool ll_sbi_has_tiny_write(struct ll_sb_info *sbi)
1095 return test_bit(LL_SBI_TINY_WRITE, sbi->ll_flags);
1098 static inline bool ll_sbi_has_file_heat(struct ll_sb_info *sbi)
1100 return test_bit(LL_SBI_FILE_HEAT, sbi->ll_flags);
1103 static inline bool ll_sbi_has_foreign_symlink(struct ll_sb_info *sbi)
1105 return test_bit(LL_SBI_FOREIGN_SYMLINK, sbi->ll_flags);
1108 static inline bool ll_sbi_has_parallel_dio(struct ll_sb_info *sbi)
1110 return test_bit(LL_SBI_PARALLEL_DIO, sbi->ll_flags);
1113 void ll_ras_enter(struct file *f, loff_t pos, size_t count);
1115 /* llite/lcommon_misc.c */
1116 int cl_ocd_update(struct obd_device *host, struct obd_device *watched,
1117 enum obd_notify_event ev, void *owner);
1118 int cl_get_grouplock(struct cl_object *obj, unsigned long gid, int nonblock,
1119 struct ll_grouplock *lg);
1120 void cl_put_grouplock(struct ll_grouplock *lg);
1122 /* llite/lproc_llite.c */
1123 int ll_debugfs_register_super(struct super_block *sb, const char *name);
1124 void ll_debugfs_unregister_super(struct super_block *sb);
1125 void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, long count);
1126 void ll_free_rw_stats_info(struct ll_sb_info *sbi);
1129 LPROC_LL_READ_BYTES,
1130 LPROC_LL_WRITE_BYTES,
1157 LPROC_LL_GETXATTR_HITS,
1159 LPROC_LL_REMOVEXATTR,
1160 LPROC_LL_INODE_PERM,
1162 LPROC_LL_INODE_OCOUNT,
1163 LPROC_LL_INODE_OPCLTM,
1164 LPROC_LL_FILE_OPCODES
1168 enum get_default_layout_type {
1169 GET_DEFAULT_LAYOUT_ROOT = 1,
1172 extern const struct file_operations ll_dir_operations;
1173 extern const struct inode_operations ll_dir_inode_operations;
1174 #ifdef HAVE_DIR_CONTEXT
1175 int ll_dir_read(struct inode *inode, __u64 *pos, struct md_op_data *op_data,
1176 struct dir_context *ctx, int *partial_readdir_rc);
1178 int ll_dir_read(struct inode *inode, __u64 *pos, struct md_op_data *op_data,
1179 void *cookie, filldir_t filldir, int *partial_readdir_rc);
1181 int ll_get_mdt_idx(struct inode *inode);
1182 int ll_get_mdt_idx_by_fid(struct ll_sb_info *sbi, const struct lu_fid *fid);
1183 struct page *ll_get_dir_page(struct inode *dir, struct md_op_data *op_data,
1184 __u64 offset, int *partial_readdir_rc);
1185 void ll_release_page(struct inode *inode, struct page *page, bool remove);
1186 int quotactl_ioctl(struct super_block *sb, struct if_quotactl *qctl);
1189 extern const struct inode_operations ll_special_inode_operations;
1191 struct inode *ll_iget(struct super_block *sb, ino_t hash,
1192 struct lustre_md *lic);
1193 int ll_test_inode_by_fid(struct inode *inode, void *opaque);
1194 int ll_md_blocking_ast(struct ldlm_lock *, struct ldlm_lock_desc *,
1195 void *data, int flag);
1196 struct dentry *ll_splice_alias(struct inode *inode, struct dentry *de);
1197 int ll_rmdir_entry(struct inode *dir, char *name, int namelen);
1198 void ll_update_times(struct ptlrpc_request *request, struct inode *inode);
1201 int ll_writepage(struct page *page, struct writeback_control *wbc);
1202 int ll_writepages(struct address_space *, struct writeback_control *wbc);
1203 int ll_readpage(struct file *file, struct page *page);
1204 #ifdef HAVE_AOPS_READ_FOLIO
1205 int ll_read_folio(struct file *file, struct folio *folio);
1207 int ll_io_read_page(const struct lu_env *env, struct cl_io *io,
1208 struct cl_page *page, struct file *file);
1209 void ll_readahead_init(struct inode *inode, struct ll_readahead_state *ras);
1210 int vvp_io_write_commit(const struct lu_env *env, struct cl_io *io);
1213 void ll_cl_add(struct inode *inode, const struct lu_env *env, struct cl_io *io,
1214 enum lcc_type type);
1215 void ll_cl_remove(struct inode *inode, const struct lu_env *env);
1216 struct ll_cl_context *ll_cl_find(struct inode *inode);
1218 extern const struct address_space_operations ll_aops;
1221 extern const struct inode_operations ll_file_inode_operations;
1222 const struct file_operations *ll_select_file_operations(struct ll_sb_info *sbi);
1223 extern int ll_have_md_lock(struct obd_export *exp, struct inode *inode,
1224 __u64 *bits, enum ldlm_mode l_req_mode);
1225 extern enum ldlm_mode ll_take_md_lock(struct inode *inode, __u64 bits,
1226 struct lustre_handle *lockh, __u64 flags,
1227 enum ldlm_mode mode);
1229 int ll_file_open(struct inode *inode, struct file *file);
1230 int ll_file_release(struct inode *inode, struct file *file);
1231 int ll_release_openhandle(struct dentry *, struct lookup_intent *);
1232 int ll_md_real_close(struct inode *inode, fmode_t fmode);
1233 void ll_track_file_opens(struct inode *inode);
1234 extern void ll_rw_stats_tally(struct ll_sb_info *sbi, pid_t pid,
1235 struct ll_file_data *file, loff_t pos,
1236 size_t count, int rw);
1237 #if defined(HAVE_USER_NAMESPACE_ARG) || defined(HAVE_INODEOPS_ENHANCED_GETATTR)
1238 int ll_getattr(struct user_namespace *mnt_userns, const struct path *path,
1239 struct kstat *stat, u32 request_mask, unsigned int flags);
1241 int ll_getattr(struct vfsmount *mnt, struct dentry *de, struct kstat *stat);
1242 #endif /* HAVE_USER_NAMESPACE_ARG */
1243 int ll_getattr_dentry(struct dentry *de, struct kstat *stat, u32 request_mask,
1244 unsigned int flags, bool foreign);
1245 #ifdef CONFIG_LUSTRE_FS_POSIX_ACL
1246 struct posix_acl *ll_get_acl(
1247 #ifdef HAVE_ACL_WITH_DENTRY
1248 struct user_namespace *, struct dentry *, int);
1249 #elif defined HAVE_GET_ACL_RCU_ARG
1250 struct inode *inode, int type, bool rcu);
1252 struct inode *inode, int type);
1253 #endif /* HAVE_GET_ACL_RCU_ARG */
1255 int ll_set_acl(struct user_namespace *mnt_userns,
1256 #ifdef HAVE_ACL_WITH_DENTRY
1257 struct dentry *dentry,
1259 struct inode *inode,
1261 struct posix_acl *acl, int type);
1262 #else /* !CONFIG_LUSTRE_FS_POSIX_ACL */
1263 #define ll_get_acl NULL
1264 #define ll_set_acl NULL
1265 #endif /* CONFIG_LUSTRE_FS_POSIX_ACL */
1267 static inline int ll_xflags_to_inode_flags(int xflags)
1269 return ((xflags & FS_XFLAG_SYNC) ? S_SYNC : 0) |
1270 ((xflags & FS_XFLAG_NOATIME) ? S_NOATIME : 0) |
1271 ((xflags & FS_XFLAG_APPEND) ? S_APPEND : 0) |
1272 ((xflags & FS_XFLAG_IMMUTABLE) ? S_IMMUTABLE : 0);
1275 static inline int ll_inode_flags_to_xflags(int inode_flags)
1277 return ((inode_flags & S_SYNC) ? FS_XFLAG_SYNC : 0) |
1278 ((inode_flags & S_NOATIME) ? FS_XFLAG_NOATIME : 0) |
1279 ((inode_flags & S_APPEND) ? FS_XFLAG_APPEND : 0) |
1280 ((inode_flags & S_IMMUTABLE) ? FS_XFLAG_IMMUTABLE : 0);
1283 int ll_migrate(struct inode *parent, struct file *file,
1284 struct lmv_user_md *lum, const char *name, __u32 flags);
1285 int ll_get_fid_by_name(struct inode *parent, const char *name,
1286 int namelen, struct lu_fid *fid, struct inode **inode);
1287 int ll_inode_permission(struct user_namespace *mnt_userns, struct inode *inode,
1289 int ll_ioctl_check_project(struct inode *inode, __u32 xflags, __u32 projid);
1290 int ll_ioctl_fsgetxattr(struct inode *inode, unsigned int cmd,
1292 int ll_ioctl_fssetxattr(struct inode *inode, unsigned int cmd,
1294 int ll_ioctl_project(struct file *file, unsigned int cmd, void __user *uarg);
1295 int ll_lov_setstripe_ea_info(struct inode *inode, struct dentry *dentry,
1296 __u64 flags, struct lov_user_md *lum,
1298 int ll_lov_getstripe_ea_info(struct inode *inode, const char *filename,
1299 struct lov_mds_md **lmm, int *lmm_size,
1300 struct ptlrpc_request **request);
1301 int ll_dir_setstripe(struct inode *inode, struct lov_user_md *lump,
1303 int ll_dir_get_default_layout(struct inode *inode, void **plmm, int *plmm_size,
1304 struct ptlrpc_request **request, u64 valid,
1305 enum get_default_layout_type type);
1306 int ll_dir_getstripe_default(struct inode *inode, void **lmmp,
1307 int *lmm_size, struct ptlrpc_request **request,
1308 struct ptlrpc_request **root_request, u64 valid);
1309 int ll_dir_getstripe(struct inode *inode, void **plmm, int *plmm_size,
1310 struct ptlrpc_request **request, u64 valid);
1311 int ll_fsync(struct file *file, loff_t start, loff_t end, int data);
1312 int ll_merge_attr(const struct lu_env *env, struct inode *inode);
1313 int ll_merge_attr_try(const struct lu_env *env, struct inode *inode);
1314 int ll_fid2path(struct inode *inode, void __user *arg);
1315 int __ll_fid2path(struct inode *inode, struct getinfo_fid2path *gfout,
1316 size_t outsize, __u32 pathlen_orig);
1317 int ll_data_version(struct inode *inode, __u64 *data_version, int flags);
1318 int ll_hsm_release(struct inode *inode);
1319 int ll_hsm_state_set(struct inode *inode, struct hsm_state_set *hss);
1320 void ll_io_set_mirror(struct cl_io *io, const struct file *file);
1322 /* llite/dcache.c */
1324 extern const struct dentry_operations ll_d_ops;
1326 bool ll_d_setup(struct dentry *de, bool do_put);
1328 static inline bool lld_is_init(struct dentry *dentry)
1330 return ll_d2d(dentry);
1333 #define ll_d_setup(de, do_put) (true)
1334 #define lld_is_init(dentry) (true)
1337 void ll_intent_drop_lock(struct lookup_intent *);
1338 void ll_intent_release(struct lookup_intent *);
1339 void ll_prune_aliases(struct inode *inode);
1340 void ll_lookup_finish_locks(struct lookup_intent *it, struct dentry *dentry);
1341 int ll_revalidate_it_finish(struct ptlrpc_request *request,
1342 struct lookup_intent *it, struct dentry *de);
1344 /* llite/llite_lib.c */
1345 extern const struct super_operations lustre_super_operations;
1347 void ll_lli_init(struct ll_inode_info *lli);
1348 int ll_fill_super(struct super_block *sb);
1349 void ll_put_super(struct super_block *sb);
1350 void ll_kill_super(struct super_block *sb);
1351 struct inode *ll_inode_from_resource_lock(struct ldlm_lock *lock);
1352 void ll_dir_clear_lsm_md(struct inode *inode);
1353 void ll_clear_inode(struct inode *inode);
1354 int volatile_ref_file(const char *volatile_name, int volatile_len,
1355 struct file **ref_file);
1356 int ll_setattr_raw(struct dentry *dentry, struct iattr *attr,
1357 enum op_xvalid xvalid, bool hsm_import);
1358 int ll_setattr(struct user_namespace *mnt_userns, struct dentry *de,
1359 struct iattr *attr);
1360 int ll_statfs(struct dentry *de, struct kstatfs *sfs);
1361 int ll_statfs_internal(struct ll_sb_info *sbi, struct obd_statfs *osfs,
1363 int ll_update_inode(struct inode *inode, struct lustre_md *md);
1364 void ll_update_inode_flags(struct inode *inode, unsigned int ext_flags);
1365 void ll_update_dir_depth_dmv(struct inode *dir, struct dentry *de);
1366 int ll_read_inode2(struct inode *inode, void *opaque);
1367 void ll_truncate_inode_pages_final(struct inode *inode);
1368 void ll_delete_inode(struct inode *inode);
1369 int ll_iocontrol(struct inode *inode, struct file *file,
1370 unsigned int cmd, void __user *uarg);
1371 int ll_flush_ctx(struct inode *inode);
1372 void ll_umount_begin(struct super_block *sb);
1373 int ll_remount_fs(struct super_block *sb, int *flags, char *data);
1374 int ll_show_options(struct seq_file *seq, struct dentry *dentry);
1375 void ll_dirty_page_discard_warn(struct inode *inode, int ioret);
1376 int ll_prep_inode(struct inode **inode, struct req_capsule *pill,
1377 struct super_block *sb, struct lookup_intent *it);
1378 int ll_obd_statfs(struct inode *inode, void __user *arg);
1379 int ll_get_max_mdsize(struct ll_sb_info *sbi, int *max_mdsize);
1380 int ll_get_default_mdsize(struct ll_sb_info *sbi, int *default_mdsize);
1381 int ll_set_default_mdsize(struct ll_sb_info *sbi, int default_mdsize);
1383 void ll_unlock_md_op_lsm(struct md_op_data *op_data);
1384 struct md_op_data *ll_prep_md_op_data(struct md_op_data *op_data,
1385 struct inode *i1, struct inode *i2,
1386 const char *name, size_t namelen,
1387 __u32 mode, enum md_op_code opc,
1389 void ll_finish_md_op_data(struct md_op_data *op_data);
1390 int ll_get_obd_name(struct inode *inode, unsigned int cmd, void __user *arg);
1391 void ll_compute_rootsquash_state(struct ll_sb_info *sbi);
1392 ssize_t ll_copy_user_md(const struct lov_user_md __user *md,
1393 struct lov_user_md **kbuf);
1394 void ll_open_cleanup(struct super_block *sb, struct req_capsule *pill);
1396 void ll_dom_finish_open(struct inode *inode, struct ptlrpc_request *req);
1398 /* Compute expected user md size when passing in a md from user space */
1399 static inline ssize_t ll_lov_user_md_size(const struct lov_user_md *lum)
1401 switch (lum->lmm_magic) {
1402 case LOV_USER_MAGIC_V1:
1403 return sizeof(struct lov_user_md_v1);
1404 case LOV_USER_MAGIC_V3:
1405 return sizeof(struct lov_user_md_v3);
1406 case LOV_USER_MAGIC_SPECIFIC:
1407 if (lum->lmm_stripe_count > LOV_MAX_STRIPE_COUNT)
1410 return lov_user_md_size(lum->lmm_stripe_count,
1411 LOV_USER_MAGIC_SPECIFIC);
1412 case LOV_USER_MAGIC_COMP_V1:
1413 return ((struct lov_comp_md_v1 *)lum)->lcm_size;
1414 case LOV_USER_MAGIC_FOREIGN:
1415 return foreign_size(lum);
1421 /* llite/llite_nfs.c */
1422 extern const struct export_operations lustre_export_operations;
1423 __u32 get_uuid2int(const char *name, int len);
1424 struct inode *search_inode_for_lustre(struct super_block *sb,
1425 const struct lu_fid *fid);
1426 int ll_dir_get_parent_fid(struct inode *dir, struct lu_fid *parent_fid);
1428 /* llite/symlink.c */
1429 extern const struct inode_operations ll_fast_symlink_inode_operations;
1432 * IO arguments for various VFS I/O interfaces.
1434 struct vvp_io_args {
1435 /** normal/sendfile/splice */
1438 struct kiocb *via_iocb;
1439 struct iov_iter *via_iter;
1449 struct ll_cl_context {
1450 struct list_head lcc_list;
1452 const struct lu_env *lcc_env;
1453 struct cl_io *lcc_io;
1454 struct cl_page *lcc_page;
1455 enum lcc_type lcc_type;
1456 pgoff_t lcc_end_index;
1459 struct ll_thread_info {
1460 struct vvp_io_args lti_args;
1461 struct ra_io_arg lti_ria;
1462 struct ll_cl_context lti_io_ctx;
1465 extern struct lu_context_key ll_thread_key;
1467 static inline struct ll_thread_info *ll_env_info(const struct lu_env *env)
1469 struct ll_thread_info *lti;
1471 lti = lu_context_key_get(&env->le_ctx, &ll_thread_key);
1472 LASSERT(lti != NULL);
1477 static inline struct vvp_io_args *ll_env_args(const struct lu_env *env)
1479 return &ll_env_info(env)->lti_args;
1482 void ll_io_init(struct cl_io *io, struct file *file, enum cl_io_type iot,
1483 struct vvp_io_args *args);
1485 /* llite/llite_mmap.c */
1487 int ll_file_mmap(struct file * file, struct vm_area_struct * vma);
1488 void policy_from_vma(union ldlm_policy_data *policy, struct vm_area_struct *vma,
1489 unsigned long addr, size_t count);
1490 struct vm_area_struct *our_vma(struct mm_struct *mm, unsigned long addr,
1493 #define ll_s2sbi(sb) (s2lsi(sb)->lsi_llsbi)
1495 /* don't need an addref as the sb_info should be holding one */
1496 static inline struct obd_export *ll_s2dtexp(struct super_block *sb)
1498 return ll_s2sbi(sb)->ll_dt_exp;
1501 /* don't need an addref as the sb_info should be holding one */
1502 static inline struct obd_export *ll_s2mdexp(struct super_block *sb)
1504 return ll_s2sbi(sb)->ll_md_exp;
1507 static inline struct client_obd *sbi2mdc(struct ll_sb_info *sbi)
1509 struct obd_device *obd = sbi->ll_md_exp->exp_obd;
1515 // FIXME: replace the name of this with LL_SB to conform to kernel stuff
1516 static inline struct ll_sb_info *ll_i2sbi(struct inode *inode)
1518 return ll_s2sbi(inode->i_sb);
1521 static inline struct obd_export *ll_i2dtexp(struct inode *inode)
1523 return ll_s2dtexp(inode->i_sb);
1526 static inline struct obd_export *ll_i2mdexp(struct inode *inode)
1528 return ll_s2mdexp(inode->i_sb);
1531 static inline struct lu_fid *ll_inode2fid(struct inode *inode)
1535 LASSERT(inode != NULL);
1536 fid = &ll_i2info(inode)->lli_fid;
1541 static inline bool ll_dir_striped(struct inode *inode)
1543 struct ll_inode_info *lli;
1547 if (!S_ISDIR(inode->i_mode))
1550 lli = ll_i2info(inode);
1551 if (!lli->lli_lsm_md)
1554 down_read(&lli->lli_lsm_sem);
1555 rc = lmv_dir_striped(lli->lli_lsm_md);
1556 up_read(&lli->lli_lsm_sem);
1561 static inline loff_t ll_file_maxbytes(struct inode *inode)
1563 struct cl_object *obj = ll_i2info(inode)->lli_clob;
1566 return MAX_LFS_FILESIZE;
1568 return min_t(loff_t, cl_object_maxbytes(obj), MAX_LFS_FILESIZE);
1572 extern const struct xattr_handler *ll_xattr_handlers[];
1574 #define XATTR_USER_T 1
1575 #define XATTR_TRUSTED_T 2
1576 #define XATTR_SECURITY_T 3
1577 #define XATTR_ACL_ACCESS_T 4
1578 #define XATTR_ACL_DEFAULT_T 5
1579 #define XATTR_LUSTRE_T 6
1580 #define XATTR_OTHER_T 7
1581 #define XATTR_ENCRYPTION_T 9
1583 ssize_t ll_listxattr(struct dentry *dentry, char *buffer, size_t size);
1584 int ll_xattr_list(struct inode *inode, const char *name, int type,
1585 void *buffer, size_t size, u64 valid);
1586 const struct xattr_handler *get_xattr_type(const char *name);
1589 * Common IO arguments for various VFS I/O interfaces.
1591 int cl_sb_init(struct super_block *sb);
1592 int cl_sb_fini(struct super_block *sb);
1594 enum ras_update_flags {
1598 void ll_ra_count_put(struct ll_sb_info *sbi, unsigned long len);
1599 void ll_ra_stats_inc(struct inode *inode, enum ra_stat which);
1603 #define LL_SA_RPC_MIN 8
1604 #define LL_SA_RPC_DEF 128
1605 #define LL_SA_RPC_MAX 2048
1607 /* XXX: If want to support more concurrent statahead instances,
1608 * please consider to decentralize the RPC lists attached
1609 * on related import, such as imp_{sending,delayed}_list.
1611 #define LL_SA_RUNNING_MAX 256
1612 #define LL_SA_RUNNING_DEF 16
1614 #define LL_SA_BATCH_MAX 1024
1615 #define LL_SA_BATCH_DEF 64
1617 #define LL_SA_CACHE_BIT 5
1618 #define LL_SA_CACHE_SIZE (1 << LL_SA_CACHE_BIT)
1619 #define LL_SA_CACHE_MASK (LL_SA_CACHE_SIZE - 1)
1621 /* per inode struct, for dir only */
1622 struct ll_statahead_info {
1623 struct dentry *sai_dentry;
1624 atomic_t sai_refcount; /* when access this struct, hold
1626 unsigned int sai_max; /* max ahead of lookup */
1627 __u64 sai_sent; /* stat requests sent count */
1628 __u64 sai_replied; /* stat requests which received
1630 __u64 sai_index; /* index of statahead entry */
1631 __u64 sai_index_wait; /* index of entry which is the
1632 * caller is waiting for */
1633 __u64 sai_hit; /* hit count */
1634 __u64 sai_miss; /* miss count:
1635 * for "ls -al" case, includes
1636 * hidden dentry miss;
1637 * for "ls -l" case, it does not
1638 * include hidden dentry miss.
1639 * "sai_miss_hidden" is used for
1642 unsigned int sai_consecutive_miss; /* consecutive miss */
1643 unsigned int sai_miss_hidden;/* "ls -al", but first dentry
1644 * is not a hidden one */
1645 unsigned int sai_skip_hidden;/* skipped hidden dentry count
1647 unsigned int sai_ls_all:1; /* "ls -al", do stat-ahead for
1649 wait_queue_head_t sai_waitq; /* stat-ahead wait queue */
1650 struct task_struct *sai_task; /* stat-ahead thread */
1651 struct task_struct *sai_agl_task; /* AGL thread */
1652 struct list_head sai_entries; /* completed entries */
1653 struct list_head sai_agls; /* AGLs to be sent */
1654 struct list_head sai_cache[LL_SA_CACHE_SIZE];
1655 spinlock_t sai_cache_lock[LL_SA_CACHE_SIZE];
1656 atomic_t sai_cache_count; /* entry count in cache */
1657 struct lu_batch *sai_bh;
1658 __u32 sai_max_batch_count;
1659 __u64 sai_index_end;
1662 int ll_revalidate_statahead(struct inode *dir, struct dentry **dentry,
1664 int ll_start_statahead(struct inode *dir, struct dentry *dentry, bool agl);
1665 void ll_authorize_statahead(struct inode *dir, void *key);
1666 void ll_deauthorize_statahead(struct inode *dir, void *key);
1669 blkcnt_t dirty_cnt(struct inode *inode);
1671 int cl_glimpse_size0(struct inode *inode, int agl);
1672 int cl_glimpse_lock(const struct lu_env *env, struct cl_io *io,
1673 struct inode *inode, struct cl_object *clob, int agl);
1675 static inline int cl_glimpse_size(struct inode *inode)
1677 return cl_glimpse_size0(inode, 0);
1680 /* AGL is 'asychronous glimpse lock', which is a speculative lock taken as
1681 * part of statahead */
1682 static inline int cl_agl(struct inode *inode)
1684 return cl_glimpse_size0(inode, 1);
1687 int ll_file_lock_ahead(struct file *file, struct llapi_lu_ladvise *ladvise);
1689 int cl_io_get(struct inode *inode, struct lu_env **envout,
1690 struct cl_io **ioout, __u16 *refcheck);
1692 static inline int ll_glimpse_size(struct inode *inode)
1694 struct ll_inode_info *lli = ll_i2info(inode);
1697 down_read(&lli->lli_glimpse_sem);
1698 rc = cl_glimpse_size(inode);
1699 lli->lli_glimpse_time = ktime_get();
1700 up_read(&lli->lli_glimpse_sem);
1704 /* dentry may statahead when statahead is enabled and current process has opened
1705 * parent directory, and this dentry hasn't accessed statahead cache before */
1707 dentry_may_statahead(struct inode *dir, struct dentry *dentry)
1709 struct ll_inode_info *lli;
1710 struct ll_dentry_data *ldd;
1712 if (ll_i2sbi(dir)->ll_sa_max == 0)
1715 lli = ll_i2info(dir);
1717 /* statahead is not allowed for this dir, there may be three causes:
1718 * 1. dir is not opened.
1719 * 2. statahead hit ratio is too low.
1720 * 3. previous stat started statahead thread failed. */
1721 if (!lli->lli_sa_enabled)
1724 /* not the same process, don't statahead */
1725 if (lli->lli_opendir_pid != current->pid)
1729 * When stating a dentry, kernel may trigger 'revalidate' or 'lookup'
1730 * multiple times, eg. for 'getattr', 'getxattr' and etc.
1731 * For patchless client, lookup intent is not accurate, which may
1732 * misguide statahead. For example:
1733 * The 'revalidate' call for 'getattr' and 'getxattr' of a dentry will
1734 * have the same intent -- IT_GETATTR, while one dentry should access
1735 * statahead cache once, otherwise statahead windows is messed up.
1736 * The solution is as following:
1737 * Assign 'lld_sa_generation' with 'lli_sa_generation' when a dentry
1738 * IT_GETATTR for the first time, and subsequent IT_GETATTR will
1739 * bypass interacting with statahead cache by checking
1740 * 'lld_sa_generation == lli->lli_sa_generation'.
1742 ldd = ll_d2d(dentry);
1743 if (ldd != NULL && lli->lli_sa_generation &&
1744 ldd->lld_sa_generation == lli->lli_sa_generation)
1750 int cl_sync_file_range(struct inode *inode, loff_t start, loff_t end,
1751 enum cl_fsync_mode mode, int ignore_layout);
1753 static inline int ll_file_nolock(const struct file *file)
1755 struct ll_file_data *fd = file->private_data;
1756 struct inode *inode = file_inode((struct file *)file);
1758 LASSERT(fd != NULL);
1759 return ((fd->fd_flags & LL_FILE_IGNORE_LOCK) ||
1760 test_bit(LL_SBI_NOLCK, ll_i2sbi(inode)->ll_flags));
1763 static inline void ll_set_lock_data(struct obd_export *exp, struct inode *inode,
1764 struct lookup_intent *it, __u64 *bits)
1766 if (!it->it_lock_set) {
1767 struct lustre_handle handle;
1769 /* If this inode is a remote object, it will get two
1770 * separate locks in different namespaces, Master MDT,
1771 * where the name entry is, will grant LOOKUP lock,
1772 * remote MDT, where the object is, will grant
1773 * UPDATE|PERM lock. The inode will be attched to both
1774 * LOOKUP and PERM locks, so revoking either locks will
1775 * case the dcache being cleared */
1776 if (it->it_remote_lock_mode) {
1777 handle.cookie = it->it_remote_lock_handle;
1778 CDEBUG(D_DLMTRACE, "setting l_data to inode "DFID
1779 "(%p) for remote lock %#llx\n",
1780 PFID(ll_inode2fid(inode)), inode,
1782 md_set_lock_data(exp, &handle, inode, NULL);
1785 handle.cookie = it->it_lock_handle;
1787 CDEBUG(D_DLMTRACE, "setting l_data to inode "DFID"(%p)"
1788 " for lock %#llx\n",
1789 PFID(ll_inode2fid(inode)), inode, handle.cookie);
1791 md_set_lock_data(exp, &handle, inode, &it->it_lock_bits);
1792 it->it_lock_set = 1;
1796 *bits = it->it_lock_bits;
1799 static inline int d_lustre_invalid(const struct dentry *dentry)
1801 return !ll_d2d(dentry) || ll_d2d(dentry)->lld_invalid;
1805 * Mark dentry INVALID, if dentry refcount is zero (this is normally case for
1806 * ll_md_blocking_ast), it will be pruned by ll_prune_aliases() and
1807 * ll_prune_negative_children(); otherwise dput() of the last refcount will
1808 * unhash this dentry and kill it.
1810 static inline void d_lustre_invalidate(struct dentry *dentry)
1812 CDEBUG(D_DENTRY, "invalidate dentry %pd (%p) parent %p inode %p refc %d\n",
1814 dentry->d_parent, dentry->d_inode, ll_d_count(dentry));
1816 spin_lock(&dentry->d_lock);
1817 if (lld_is_init(dentry))
1818 ll_d2d(dentry)->lld_invalid = 1;
1819 spin_unlock(&dentry->d_lock);
1822 static inline void d_lustre_revalidate(struct dentry *dentry)
1824 spin_lock(&dentry->d_lock);
1825 LASSERT(ll_d2d(dentry));
1826 ll_d2d(dentry)->lld_invalid = 0;
1827 spin_unlock(&dentry->d_lock);
1830 static inline dev_t ll_compat_encode_dev(dev_t dev)
1832 /* The compat_sys_*stat*() syscalls will fail unless the
1833 * device majors and minors are both less than 256. Note that
1834 * the value returned here will be passed through
1835 * old_encode_dev() in cp_compat_stat(). And so we are not
1836 * trying to return a valid compat (u16) device number, just
1837 * one that will pass the old_valid_dev() check. */
1839 return MKDEV(MAJOR(dev) & 0xff, MINOR(dev) & 0xff);
1842 int ll_layout_conf(struct inode *inode, const struct cl_object_conf *conf);
1843 int ll_layout_refresh(struct inode *inode, __u32 *gen);
1844 int ll_layout_restore(struct inode *inode, loff_t start, __u64 length);
1845 int ll_layout_write_intent(struct inode *inode, enum layout_intent_opc opc,
1846 struct lu_extent *ext);
1848 int ll_xattr_init(void);
1849 void ll_xattr_fini(void);
1851 int ll_page_sync_io(const struct lu_env *env, struct cl_io *io,
1852 struct cl_page *page, enum cl_req_type crt);
1854 int ll_getparent(struct file *file, struct getparent __user *arg);
1857 int cl_setattr_ost(struct cl_object *obj, const struct iattr *attr,
1858 enum op_xvalid xvalid, unsigned int attr_flags);
1860 extern struct lu_env *cl_inode_fini_env;
1861 extern __u16 cl_inode_fini_refcheck;
1863 int cl_file_inode_init(struct inode *inode, struct lustre_md *md);
1864 void cl_inode_fini(struct inode *inode);
1866 u64 cl_fid_build_ino(const struct lu_fid *fid, int api32);
1867 u32 cl_fid_build_gen(const struct lu_fid *fid);
1869 static inline struct pcc_super *ll_i2pccs(struct inode *inode)
1871 return &ll_i2sbi(inode)->ll_pcc_super;
1874 static inline struct pcc_super *ll_info2pccs(struct ll_inode_info *lli)
1876 return ll_i2pccs(ll_info2i(lli));
1880 /* The digested form is made of a FID (16 bytes) followed by the second-to-last
1881 * ciphertext block (16 bytes), so a total length of 32 bytes.
1882 * That way, llcrypt does not compute a digested form of this digest.
1884 struct ll_digest_filename {
1885 struct lu_fid ldf_fid;
1886 char ldf_excerpt[LL_CRYPTO_BLOCK_SIZE];
1889 int ll_prepare_lookup(struct inode *dir, struct dentry *de,
1890 struct llcrypt_name *fname, struct lu_fid *fid);
1891 int ll_setup_filename(struct inode *dir, const struct qstr *iname,
1892 int lookup, struct llcrypt_name *fname,
1893 struct lu_fid *fid);
1894 int ll_fname_disk_to_usr(struct inode *inode,
1895 u32 hash, u32 minor_hash,
1896 struct llcrypt_str *iname, struct llcrypt_str *oname,
1897 struct lu_fid *fid);
1898 int ll_file_open_encrypt(struct inode *inode, struct file *filp);
1900 static inline char *xattr_for_enc(struct inode *inode)
1902 if (ll_sbi_has_name_encrypt(ll_i2sbi(inode)))
1903 return LL_XATTR_NAME_ENCRYPTION_CONTEXT;
1905 return LL_XATTR_NAME_ENCRYPTION_CONTEXT_OLD;
1908 #ifdef HAVE_LUSTRE_CRYPTO
1909 extern const struct llcrypt_operations lustre_cryptops;
1912 /* llite/llite_foreign.c */
1913 int ll_manage_foreign(struct inode *inode, struct lustre_md *lmd);
1914 bool ll_foreign_is_openable(struct dentry *dentry, unsigned int flags);
1915 bool ll_foreign_is_removable(struct dentry *dentry, bool unset);
1917 int ll_filemap_fault(struct vm_area_struct *vma, struct vm_fault *vmf);
1919 #endif /* LLITE_INTERNAL_H */