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 char *lli_symlink_name;
210 struct ll_trunc_sem lli_trunc_sem;
211 struct range_lock_tree lli_write_tree;
212 struct mutex lli_setattr_mutex;
214 struct rw_semaphore lli_glimpse_sem;
215 ktime_t lli_glimpse_time;
216 struct list_head lli_agl_list;
219 /* for writepage() only to communicate to fsync */
222 /* protect the file heat fields */
223 spinlock_t lli_heat_lock;
224 __u32 lli_heat_flags;
225 struct obd_heat_instance lli_heat_instances[OBD_HEAT_COUNT];
228 * Whenever a process try to read/write the file, the
229 * jobid, uid and gid of the process will be saved here,
230 * and it'll be packed into write RPC when flush later.
232 * So the read/write statistics or TBF rules for jobid,
233 * uid or gid will not be accurate if the file is shared
236 char lli_jobid[LUSTRE_JOBID_SIZE];
240 struct mutex lli_pcc_lock;
241 enum lu_pcc_state_flags lli_pcc_state;
243 * @lli_pcc_generation saves the gobal PCC generation
244 * when the file was successfully attached into PCC.
245 * The flags of the PCC dataset are saved in
247 * The gobal PCC generation will be increased when add
248 * or delete a PCC backend, or change the configuration
249 * parameters for PCC.
250 * If @lli_pcc_generation is same as the gobal PCC
251 * generation, we can use the saved flags of the PCC
252 * dataset to determine whether need to try auto attach
255 __u64 lli_pcc_generation;
256 enum pcc_dataset_flags lli_pcc_dsflags;
257 struct pcc_inode *lli_pcc_inode;
259 struct mutex lli_group_mutex;
260 __u64 lli_group_users;
261 unsigned long lli_group_gid;
263 __u64 lli_attr_valid;
265 __u64 lli_lazyblocks;
269 /* XXX: For following frequent used members, although they maybe special
270 * used for non-directory object, it is some time-wasting to check
271 * whether the object is directory or not before using them. On the
272 * other hand, currently, sizeof(f) > sizeof(d), it cannot reduce
273 * the "ll_inode_info" size even if moving those members into u.f.
274 * So keep them out side.
276 * In the future, if more members are added only for directory,
277 * some of the following members can be moved into u.f.
279 struct cl_object *lli_clob;
281 /* mutex to request for layout lock exclusively. */
282 struct mutex lli_layout_mutex;
283 /* Layout version, protected by lli_layout_lock */
284 __u32 lli_layout_gen;
285 spinlock_t lli_layout_lock;
287 __u32 lli_projid; /* project id */
289 struct rw_semaphore lli_xattrs_list_rwsem;
290 struct mutex lli_xattrs_enq_lock;
291 struct list_head lli_xattrs; /* ll_xattr_entry->xe_list */
292 struct list_head lli_lccs; /* list of ll_cl_context */
293 seqlock_t lli_page_inv_lock;
296 #ifndef HAVE_USER_NAMESPACE_ARG
297 #define inode_permission(ns, inode, mask) inode_permission(inode, mask)
298 #define generic_permission(ns, inode, mask) generic_permission(inode, mask)
299 #define simple_setattr(ns, de, iattr) simple_setattr(de, iattr)
300 #define ll_inode_permission(ns, inode, mask) ll_inode_permission(inode, mask)
301 #ifdef HAVE_INODEOPS_ENHANCED_GETATTR
302 #define ll_getattr(ns, path, stat, mask, fl) ll_getattr(path, stat, mask, fl)
303 #endif /* HAVE_INODEOPS_ENHANCED_GETATTR */
304 #define ll_setattr(ns, de, attr) ll_setattr(de, attr)
307 static inline void ll_trunc_sem_init(struct ll_trunc_sem *sem)
309 atomic_set(&sem->ll_trunc_readers, 0);
310 atomic_set(&sem->ll_trunc_waiters, 0);
313 /* This version of down read ignores waiting writers, meaning if the semaphore
314 * is already held for read, this down_read will 'join' that reader and also
315 * take the semaphore.
317 * This lets us avoid an unusual deadlock.
319 * We must take lli_trunc_sem in read mode on entry in to various i/o paths
320 * in Lustre, in order to exclude truncates. Some of these paths then need to
321 * take the mmap_lock, while still holding the trunc_sem. The problem is that
322 * page faults hold the mmap_lock when calling in to Lustre, and then must also
323 * take the trunc_sem to exclude truncate.
325 * This means the locking order for trunc_sem and mmap_lock is sometimes AB,
326 * sometimes BA. This is almost OK because in both cases, we take the trunc
327 * sem for read, so it doesn't block.
329 * However, if a write mode user (truncate, a setattr op) arrives in the
330 * middle of this, the second reader on the truncate_sem will wait behind that
333 * So we have, on our truncate sem, in order (where 'reader' and 'writer' refer
334 * to the mode in which they take the semaphore):
335 * reader (holding mmap_lock, needs truncate_sem)
337 * reader (holding truncate sem, waiting for mmap_lock)
339 * And so the readers deadlock.
341 * The solution is this modified semaphore, where this down_read ignores
342 * waiting write operations, and all waiters are woken up at once, so readers
343 * using down_read_nowait cannot get stuck behind waiting writers, regardless
344 * of the order they arrived in.
346 * down_read_nowait is only used in the page fault case, where we already hold
347 * the mmap_lock. This is because otherwise repeated read and write operations
348 * (which take the truncate sem) could prevent a truncate from ever starting.
349 * This could still happen with page faults, but without an even more complex
350 * mechanism, this is unavoidable.
354 static inline void trunc_sem_down_read_nowait(struct ll_trunc_sem *sem)
356 wait_var_event(&sem->ll_trunc_readers,
357 atomic_inc_unless_negative(&sem->ll_trunc_readers));
360 static inline void trunc_sem_down_read(struct ll_trunc_sem *sem)
362 wait_var_event(&sem->ll_trunc_readers,
363 atomic_read(&sem->ll_trunc_waiters) == 0 &&
364 atomic_inc_unless_negative(&sem->ll_trunc_readers));
367 static inline void trunc_sem_up_read(struct ll_trunc_sem *sem)
369 if (atomic_dec_return(&sem->ll_trunc_readers) == 0 &&
370 atomic_read(&sem->ll_trunc_waiters))
371 wake_up_var(&sem->ll_trunc_readers);
374 static inline void trunc_sem_down_write(struct ll_trunc_sem *sem)
376 atomic_inc(&sem->ll_trunc_waiters);
377 wait_var_event(&sem->ll_trunc_readers,
378 atomic_cmpxchg(&sem->ll_trunc_readers, 0, -1) == 0);
379 atomic_dec(&sem->ll_trunc_waiters);
382 static inline void trunc_sem_up_write(struct ll_trunc_sem *sem)
384 atomic_set(&sem->ll_trunc_readers, 0);
385 /* match the smp_mb() in wait_var_event()->prepare_to_wait() */
387 wake_up_var(&sem->ll_trunc_readers);
390 #ifdef CONFIG_LUSTRE_FS_POSIX_ACL
391 static inline void lli_clear_acl(struct ll_inode_info *lli)
393 if (lli->lli_posix_acl) {
394 posix_acl_release(lli->lli_posix_acl);
395 lli->lli_posix_acl = NULL;
399 static inline void lli_replace_acl(struct ll_inode_info *lli,
400 struct lustre_md *md)
402 write_lock(&lli->lli_lock);
403 if (lli->lli_posix_acl)
404 posix_acl_release(lli->lli_posix_acl);
405 lli->lli_posix_acl = md->posix_acl;
406 write_unlock(&lli->lli_lock);
409 static inline void lli_clear_acl(struct ll_inode_info *lli)
413 static inline void lli_replace_acl(struct ll_inode_info *lli,
414 struct lustre_md *md)
419 static inline __u32 ll_layout_version_get(struct ll_inode_info *lli)
423 spin_lock(&lli->lli_layout_lock);
424 gen = lli->lli_layout_gen;
425 spin_unlock(&lli->lli_layout_lock);
430 static inline void ll_layout_version_set(struct ll_inode_info *lli, __u32 gen)
432 spin_lock(&lli->lli_layout_lock);
433 lli->lli_layout_gen = gen;
434 spin_unlock(&lli->lli_layout_lock);
438 /* File data is modified. */
439 LLIF_DATA_MODIFIED = 0,
440 /* File is being restored */
441 LLIF_FILE_RESTORING = 1,
442 /* Xattr cache is attached to the file */
443 LLIF_XATTR_CACHE = 2,
444 /* Project inherit */
445 LLIF_PROJECT_INHERIT = 3,
446 /* update atime from MDS even if it's older than local inode atime. */
447 LLIF_UPDATE_ATIME = 4,
448 /* foreign file/dir can be unlinked unconditionnaly */
449 LLIF_FOREIGN_REMOVABLE = 5,
450 /* Xattr cache is filled */
451 LLIF_XATTR_CACHE_FILLED = 7,
455 int ll_xattr_cache_destroy(struct inode *inode);
456 int ll_xattr_cache_empty(struct inode *inode);
458 int ll_xattr_cache_get(struct inode *inode,
464 int ll_xattr_cache_insert(struct inode *inode,
469 static inline bool obd_connect_has_secctx(struct obd_connect_data *data)
471 #ifdef CONFIG_SECURITY
472 return data->ocd_connect_flags & OBD_CONNECT_FLAGS2 &&
473 data->ocd_connect_flags2 & OBD_CONNECT2_FILE_SECCTX;
479 static inline void obd_connect_set_secctx(struct obd_connect_data *data)
481 #ifdef CONFIG_SECURITY
482 data->ocd_connect_flags2 |= OBD_CONNECT2_FILE_SECCTX;
486 /* Only smack and selinux is known to use security contexts */
487 static inline bool ll_xattr_is_seclabel(const char *name)
489 return !strcmp(name, XATTR_NAME_SELINUX) ||
490 !strcmp(name, XATTR_NAME_SMACK);
493 static inline bool ll_xattr_suffix_is_seclabel(const char *suffix)
495 return !strcmp(suffix, XATTR_SELINUX_SUFFIX) ||
496 !strcmp(suffix, XATTR_SMACK_SUFFIX);
499 int ll_dentry_init_security(struct dentry *dentry, int mode, struct qstr *name,
500 const char **secctx_name, __u32 *secctx_name_size,
501 void **secctx, __u32 *secctx_size,
504 int ll_inode_init_security(struct dentry *dentry, struct inode *inode,
507 int ll_inode_notifysecctx(struct inode *inode,
508 void *secctx, __u32 secctxlen);
510 void ll_secctx_name_free(struct ll_sb_info *sbi);
512 int ll_secctx_name_store(struct inode *in);
514 __u32 ll_secctx_name_get(struct ll_sb_info *sbi, const char **secctx_name);
516 int ll_security_secctx_name_filter(struct ll_sb_info *sbi, int xattr_type,
519 static inline bool obd_connect_has_enc(struct obd_connect_data *data)
521 #ifdef HAVE_LUSTRE_CRYPTO
522 return data->ocd_connect_flags & OBD_CONNECT_FLAGS2 &&
523 data->ocd_connect_flags2 & OBD_CONNECT2_ENCRYPT;
529 static inline void obd_connect_set_enc(struct obd_connect_data *data)
531 #ifdef HAVE_LUSTRE_CRYPTO
532 data->ocd_connect_flags2 |= OBD_CONNECT2_ENCRYPT;
536 static inline bool obd_connect_has_name_enc(struct obd_connect_data *data)
538 #ifdef HAVE_LUSTRE_CRYPTO
539 return data->ocd_connect_flags & OBD_CONNECT_FLAGS2 &&
540 data->ocd_connect_flags2 & OBD_CONNECT2_ENCRYPT_NAME;
546 static inline void obd_connect_set_name_enc(struct obd_connect_data *data)
548 #ifdef HAVE_LUSTRE_CRYPTO
549 data->ocd_connect_flags2 |= OBD_CONNECT2_ENCRYPT_NAME;
553 static inline bool obd_connect_has_enc_fid2path(struct obd_connect_data *data)
555 #ifdef HAVE_LUSTRE_CRYPTO
556 return data->ocd_connect_flags & OBD_CONNECT_FLAGS2 &&
557 data->ocd_connect_flags2 & OBD_CONNECT2_ENCRYPT_FID2PATH;
563 static inline void obd_connect_set_enc_fid2path(struct obd_connect_data *data)
565 #ifdef HAVE_LUSTRE_CRYPTO
566 data->ocd_connect_flags2 |= OBD_CONNECT2_ENCRYPT_FID2PATH;
571 * Locking to guarantee consistency of non-atomic updates to long long i_size,
572 * consistency between file size and KMS.
574 * Implemented by ->lli_size_mutex and ->lsm_lock, nested in that order.
577 void ll_inode_size_lock(struct inode *inode);
578 void ll_inode_size_unlock(struct inode *inode);
579 int ll_inode_size_trylock(struct inode *inode);
581 static inline struct ll_inode_info *ll_i2info(struct inode *inode)
583 return container_of(inode, struct ll_inode_info, lli_vfs_inode);
586 static inline struct pcc_inode *ll_i2pcci(struct inode *inode)
588 return ll_i2info(inode)->lli_pcc_inode;
591 /* default to use at least 16M for fast read if possible */
592 #define RA_REMAIN_WINDOW_MIN MiB_TO_PAGES(16UL)
594 /* default read-ahead on a given client mountpoint. */
595 #define SBI_DEFAULT_READ_AHEAD_MAX MiB_TO_PAGES(1024UL)
597 /* default read-ahead for a single file descriptor */
598 #define SBI_DEFAULT_READ_AHEAD_PER_FILE_MAX MiB_TO_PAGES(256UL)
600 /* default read-ahead full files smaller than limit on the second read */
601 #define SBI_DEFAULT_READ_AHEAD_WHOLE_MAX MiB_TO_PAGES(2UL)
603 /* default range pages */
604 #define SBI_DEFAULT_RA_RANGE_PAGES MiB_TO_PAGES(1ULL)
606 /* Min range pages */
607 #define RA_MIN_MMAP_RANGE_PAGES 16UL
612 RA_STAT_DISTANT_READPAGE,
613 RA_STAT_MISS_IN_WINDOW,
614 RA_STAT_FAILED_GRAB_PAGE,
615 RA_STAT_FAILED_MATCH,
620 RA_STAT_MAX_IN_FLIGHT,
621 RA_STAT_WRONG_GRAB_PAGE,
622 RA_STAT_FAILED_REACH_END,
624 RA_STAT_FAILED_FAST_READ,
625 RA_STAT_MMAP_RANGE_READ,
626 RA_STAT_READAHEAD_PAGES,
631 atomic_t ra_cur_pages;
632 unsigned long ra_max_pages;
633 unsigned long ra_max_pages_per_file;
634 unsigned long ra_range_pages;
635 unsigned long ra_max_read_ahead_whole_pages;
636 struct workqueue_struct *ll_readahead_wq;
638 * Max number of active works could be triggered
639 * for async readahead.
641 unsigned int ra_async_max_active;
642 /* how many async readahead triggered in flight */
643 atomic_t ra_async_inflight;
644 /* Threshold to control when to trigger async readahead */
645 unsigned long ra_async_pages_per_file_threshold;
648 /* ra_io_arg will be filled in the beginning of ll_readahead with
649 * ras_lock, then the following ll_read_ahead_pages will read RA
650 * pages according to this arg, all the items in this structure are
651 * counted by page index.
654 pgoff_t ria_start_idx; /* start offset of read-ahead*/
655 pgoff_t ria_end_idx; /* end offset of read-ahead*/
656 unsigned long ria_reserved; /* reserved pages for read-ahead */
657 pgoff_t ria_end_idx_min;/* minimum end to cover current read */
658 bool ria_eof; /* reach end of file */
659 /* If stride read pattern is detected, ria_stoff is the byte offset
660 * where stride read is started. Note: for normal read-ahead, the
661 * value here is meaningless, and also it will not be accessed*/
663 /* ria_length and ria_bytes are the length and pages length in the
664 * stride I/O mode. And they will also be used to check whether
665 * it is stride I/O read-ahead in the read-ahead pages*/
670 /* LL_HIST_MAX=32 causes an overflow */
671 #define LL_HIST_MAX 28
672 #define LL_HIST_START 12 /* buckets start at 2^12 = 4k */
673 #define LL_PROCESS_HIST_MAX 10
674 struct per_process_info {
676 struct obd_histogram pp_r_hist;
677 struct obd_histogram pp_w_hist;
680 /* pp_extents[LL_PROCESS_HIST_MAX] will hold the combined process info */
681 struct ll_rw_extents_info {
683 struct per_process_info pp_extents[LL_PROCESS_HIST_MAX + 1];
686 #define LL_OFFSET_HIST_MAX 100
687 struct ll_rw_process_info {
690 loff_t rw_range_start;
692 loff_t rw_last_file_pos;
694 size_t rw_smallest_extent;
695 size_t rw_largest_extent;
696 struct ll_file_data *rw_last_file;
699 enum stats_track_type {
700 STATS_TRACK_ALL = 0, /* track all processes */
701 STATS_TRACK_PID, /* track process with this pid */
702 STATS_TRACK_PPID, /* track processes with this ppid */
703 STATS_TRACK_GID, /* track processes with this gid */
707 /* flags for sbi->ll_flags */
709 LL_SBI_NOLCK, /* DLM locking disabled directio-only */
710 LL_SBI_CHECKSUM, /* checksum each page as it's written */
711 LL_SBI_LOCALFLOCK, /* local flocks instead of fs-wide */
712 LL_SBI_FLOCK, /* flock enabled */
713 LL_SBI_USER_XATTR, /* support user xattr */
714 LL_SBI_LRU_RESIZE, /* lru resize support */
715 LL_SBI_LAZYSTATFS, /* lazystatfs mount option */
716 LL_SBI_32BIT_API, /* generate 32 bit inodes. */
717 LL_SBI_USER_FID2PATH, /* fid2path by unprivileged users */
718 LL_SBI_VERBOSE, /* verbose mount/umount */
719 LL_SBI_ALWAYS_PING, /* ping even if server suppress_pings */
720 LL_SBI_TEST_DUMMY_ENCRYPTION, /* test dummy encryption */
721 LL_SBI_ENCRYPT, /* client side encryption */
722 LL_SBI_FOREIGN_SYMLINK, /* foreign fake-symlink support */
723 LL_SBI_FOREIGN_SYMLINK_UPCALL, /* foreign fake-symlink upcall set */
724 LL_SBI_NUM_MOUNT_OPT,
726 LL_SBI_ACL, /* support ACL */
727 LL_SBI_AGL_ENABLED, /* enable agl */
728 LL_SBI_64BIT_HASH, /* support 64-bits dir hash/offset */
729 LL_SBI_LAYOUT_LOCK, /* layout lock support */
730 LL_SBI_XATTR_CACHE, /* support for xattr cache */
731 LL_SBI_NOROOTSQUASH, /* do not apply root squash */
732 LL_SBI_FAST_READ, /* fast read support */
733 LL_SBI_FILE_SECCTX, /* file security context at create */
734 LL_SBI_TINY_WRITE, /* tiny write support */
735 LL_SBI_FILE_HEAT, /* file heat support */
736 LL_SBI_PARALLEL_DIO, /* parallel (async) O_DIRECT RPCs */
737 LL_SBI_ENCRYPT_NAME, /* name encryption */
741 int ll_sbi_flags_seq_show(struct seq_file *m, void *v);
743 /* This is embedded into llite super-blocks to keep track of connect
744 * flags (capabilities) supported by all imports given mount is
746 struct lustre_client_ocd {
747 /* This is conjunction of connect_flags across all imports
748 * (LOVs) this mount is connected to. This field is updated by
749 * cl_ocd_update() under ->lco_lock. */
751 struct mutex lco_lock;
752 struct obd_export *lco_md_exp;
753 struct obd_export *lco_dt_exp;
757 /* this protects pglist and ra_info. It isn't safe to
758 * grab from interrupt contexts */
760 spinlock_t ll_pp_extent_lock; /* pp_extent entry*/
761 spinlock_t ll_process_lock; /* ll_rw_process_info */
762 struct lustre_sb_info *lsi;
763 struct obd_uuid ll_sb_uuid;
764 struct obd_export *ll_md_exp;
765 struct obd_export *ll_dt_exp;
766 struct obd_device *ll_md_obd;
767 struct obd_device *ll_dt_obd;
768 struct dentry *ll_debugfs_entry;
769 struct lu_fid ll_root_fid; /* root object fid */
770 struct mnt_namespace *ll_mnt_ns;
772 DECLARE_BITMAP(ll_flags, LL_SBI_NUM_FLAGS); /* enum ll_sbi_flags */
773 unsigned int ll_xattr_cache_enabled:1,
774 ll_xattr_cache_set:1, /* already set to 0/1 */
775 ll_client_common_fill_super_succeeded:1,
777 ll_inode_cache_enabled:1;
779 struct lustre_client_ocd ll_lco;
781 struct lprocfs_stats *ll_stats; /* lprocfs stats counter */
783 /* Used to track "unstable" pages on a client, and maintain a
784 * LRU list of clean pages. An "unstable" page is defined as
785 * any page which is sent to a server as part of a bulk request,
786 * but is uncommitted to stable storage. */
787 struct cl_client_cache *ll_cache;
789 struct lprocfs_stats *ll_ra_stats;
791 struct ll_ra_info ll_ra_info;
792 unsigned int ll_namelen;
793 const struct file_operations *ll_fop;
795 struct lu_site *ll_site;
796 struct cl_device *ll_cl;
799 struct ll_rw_extents_info *ll_rw_extents_info;
800 int ll_extent_process_count;
801 unsigned int ll_offset_process_count;
802 struct ll_rw_process_info *ll_rw_process_info;
803 struct ll_rw_process_info *ll_rw_offset_info;
804 ktime_t ll_process_stats_init;
805 unsigned int ll_rw_offset_entry_count;
806 int ll_stats_track_id;
807 enum stats_track_type ll_stats_track_type;
810 /* metadata stat-ahead */
811 unsigned int ll_sa_running_max;/* max concurrent
812 * statahead instances */
813 unsigned int ll_sa_batch_max;/* max SUB request count in
814 * a batch PTLRPC request */
815 unsigned int ll_sa_max; /* max statahead RPCs */
816 atomic_t ll_sa_total; /* statahead thread started
818 atomic_t ll_sa_wrong; /* statahead thread stopped for
820 atomic_t ll_sa_running; /* running statahead thread
822 atomic_t ll_agl_total; /* AGL thread started count */
823 atomic_t ll_sa_hit_total; /* total hit count */
824 atomic_t ll_sa_miss_total; /* total miss count */
826 dev_t ll_sdev_orig; /* save s_dev before assign for
829 struct root_squash_info ll_squash;
832 /* st_blksize returned by stat(2), when non-zero */
833 unsigned int ll_stat_blksize;
835 /* maximum relative age of cached statfs results */
836 unsigned int ll_statfs_max_age;
838 struct kset ll_kset; /* sysfs object */
839 struct completion ll_kobj_unregister;
842 unsigned int ll_heat_decay_weight;
843 unsigned int ll_heat_period_second;
845 /* Opens of the same inode before we start requesting open lock */
846 u32 ll_oc_thrsh_count;
848 /* Time in ms between last inode close and next open to be considered
849 * instant back to back and would trigger an open lock request
853 /* Time in ms after last file close that we no longer count prior opens*/
856 /* filesystem fsname */
857 char ll_fsname[LUSTRE_MAXFSNAME + 1];
859 /* Persistent Client Cache */
860 struct pcc_super ll_pcc_super;
862 /* to protect vs updates in all following foreign symlink fields */
863 struct rw_semaphore ll_foreign_symlink_sem;
864 /* foreign symlink path prefix */
865 char *ll_foreign_symlink_prefix;
866 /* full prefix size including leading '\0' */
867 size_t ll_foreign_symlink_prefix_size;
868 /* foreign symlink path upcall */
869 char *ll_foreign_symlink_upcall;
870 /* foreign symlink path upcall infos */
871 struct ll_foreign_symlink_upcall_item *ll_foreign_symlink_upcall_items;
872 /* foreign symlink path upcall nb infos */
873 unsigned int ll_foreign_symlink_upcall_nb_items;
875 /* cached file security context xattr name. e.g: security.selinux */
876 char *ll_secctx_name;
877 __u32 ll_secctx_name_size;
880 #define SBI_DEFAULT_HEAT_DECAY_WEIGHT ((80 * 256 + 50) / 100)
881 #define SBI_DEFAULT_HEAT_PERIOD_SECOND (60)
883 #define SBI_DEFAULT_OPENCACHE_THRESHOLD_COUNT (5)
884 #define SBI_DEFAULT_OPENCACHE_THRESHOLD_MS (100) /* 0.1 second */
885 #define SBI_DEFAULT_OPENCACHE_THRESHOLD_MAX_MS (60000) /* 1 minute */
888 * per file-descriptor read-ahead data.
890 struct ll_readahead_state {
892 /* End byte that read(2) try to read. */
893 loff_t ras_last_read_end_bytes;
895 * number of bytes read after last read-ahead window reset. As window
896 * is reset on each seek, this is effectively a number of consecutive
897 * accesses. Maybe ->ras_accessed_in_window is better name.
899 * XXX nikita: window is also reset (by ras_update()) when Lustre
900 * believes that memory pressure evicts read-ahead pages. In that
901 * case, it probably doesn't make sense to expand window to
902 * PTLRPC_MAX_BRW_PAGES on the third access.
904 loff_t ras_consecutive_bytes;
906 * number of read requests after the last read-ahead window reset
907 * As window is reset on each seek, this is effectively the number
908 * on consecutive read request and is used to trigger read-ahead.
910 unsigned long ras_consecutive_requests;
912 * Parameters of current read-ahead window. Handled by
913 * ras_update(). On the initial access to the file or after a seek,
914 * window is reset to 0. After 3 consecutive accesses, window is
915 * expanded to PTLRPC_MAX_BRW_PAGES. Afterwards, window is enlarged by
916 * PTLRPC_MAX_BRW_PAGES chunks up to ->ra_max_pages.
918 pgoff_t ras_window_start_idx;
919 pgoff_t ras_window_pages;
921 /* Page index where min range read starts */
922 pgoff_t ras_range_min_start_idx;
923 /* Page index where mmap range read ends */
924 pgoff_t ras_range_max_end_idx;
925 /* number of mmap pages where last time detected */
926 pgoff_t ras_last_range_pages;
927 /* number of mmap range requests */
928 pgoff_t ras_range_requests;
931 * Optimal RPC size in pages.
932 * It decides how many pages will be sent for each read-ahead.
934 unsigned long ras_rpc_pages;
936 * Where next read-ahead should start at. This lies within read-ahead
937 * window. Read-ahead window is read in pieces rather than at once
938 * because: 1. lustre limits total number of pages under read-ahead by
939 * ->ra_max_pages (see ll_ra_count_get()), 2. client cannot read pages
940 * not covered by DLM lock.
942 pgoff_t ras_next_readahead_idx;
944 * Total number of ll_file_read requests issued, reads originating
945 * due to mmap are not counted in this total. This value is used to
946 * trigger full file read-ahead after multiple reads to a small file.
948 unsigned long ras_requests;
950 * The following 3 items are used for detecting the stride I/O
952 * In stride I/O mode,
953 * ...............|-----data-----|****gap*****|--------|******|....
954 * offset |-stride_bytes-|-stride_gap-|
955 * ras_stride_offset = offset;
956 * ras_stride_length = stride_bytes + stride_gap;
957 * ras_stride_bytes = stride_bytes;
958 * Note: all these three items are counted by bytes.
960 loff_t ras_stride_offset;
961 loff_t ras_stride_length;
962 loff_t ras_stride_bytes;
964 * number of consecutive stride request count, and it is similar as
965 * ras_consecutive_requests, but used for stride I/O mode.
966 * Note: only more than 2 consecutive stride request are detected,
967 * stride read-ahead will be enable
969 unsigned long ras_consecutive_stride_requests;
970 /* index of the last page that async readahead starts */
971 pgoff_t ras_async_last_readpage_idx;
972 /* whether we should increase readahead window */
973 bool ras_need_increase_window;
974 /* whether ra miss check should be skipped */
975 bool ras_no_miss_check;
978 struct ll_readahead_work {
979 /** File to readahead */
980 struct file *lrw_file;
981 pgoff_t lrw_start_idx;
985 /* async worker to handler read */
986 struct work_struct lrw_readahead_work;
987 char lrw_jobid[LUSTRE_JOBID_SIZE];
990 extern struct kmem_cache *ll_file_data_slab;
991 struct lustre_handle;
992 struct ll_file_data {
993 struct ll_readahead_state fd_ras;
994 struct ll_grouplock fd_grouplock;
998 /* openhandle if lease exists for this file.
999 * Borrow lli->lli_och_mutex to protect assignment */
1000 struct obd_client_handle *fd_lease_och;
1001 struct obd_client_handle *fd_och;
1002 struct file *fd_file;
1003 /* Indicate whether need to report failure when close.
1004 * true: failure is known, not report again.
1005 * false: unknown failure, should report. */
1006 bool fd_write_failed;
1007 bool ll_lock_no_expand;
1008 /* Used by mirrored file to lead IOs to a specific mirror, usually
1009 * for mirror resync. 0 means default. */
1010 __u32 fd_designated_mirror;
1011 /* The layout version when resync starts. Resync I/O should carry this
1012 * layout version for verification to OST objects */
1013 __u32 fd_layout_version;
1014 struct pcc_file fd_pcc_file;
1015 /* striped directory may read partially if some stripe inaccessible,
1016 * -errno is saved here, and will return to user in close().
1018 int fd_partial_readdir_rc;
1021 void llite_tunables_unregister(void);
1022 int llite_tunables_register(void);
1024 static inline struct inode *ll_info2i(struct ll_inode_info *lli)
1026 return &lli->lli_vfs_inode;
1029 __u32 ll_i2suppgid(struct inode *i);
1030 void ll_i2gids(__u32 *suppgids, struct inode *i1,struct inode *i2);
1032 static inline int ll_need_32bit_api(struct ll_sb_info *sbi)
1034 #if BITS_PER_LONG == 32
1036 #elif defined(CONFIG_COMPAT)
1037 if (unlikely(test_bit(LL_SBI_32BIT_API, sbi->ll_flags)))
1040 # ifdef CONFIG_X86_X32
1041 /* in_compat_syscall() returns true when called from a kthread
1042 * and CONFIG_X86_X32 is enabled, which is wrong. So check
1043 * whether the caller comes from a syscall (ie. not a kthread)
1044 * before calling in_compat_syscall(). */
1045 if (current->flags & PF_KTHREAD)
1049 return unlikely(in_compat_syscall());
1051 return unlikely(test_bit(LL_SBI_32BIT_API, sbi->ll_flags));
1055 static inline bool ll_sbi_has_fast_read(struct ll_sb_info *sbi)
1057 return test_bit(LL_SBI_FAST_READ, sbi->ll_flags);
1060 static inline bool ll_sbi_has_tiny_write(struct ll_sb_info *sbi)
1062 return test_bit(LL_SBI_TINY_WRITE, sbi->ll_flags);
1065 static inline bool ll_sbi_has_file_heat(struct ll_sb_info *sbi)
1067 return test_bit(LL_SBI_FILE_HEAT, sbi->ll_flags);
1070 static inline bool ll_sbi_has_foreign_symlink(struct ll_sb_info *sbi)
1072 return test_bit(LL_SBI_FOREIGN_SYMLINK, sbi->ll_flags);
1075 static inline bool ll_sbi_has_parallel_dio(struct ll_sb_info *sbi)
1077 return test_bit(LL_SBI_PARALLEL_DIO, sbi->ll_flags);
1080 void ll_ras_enter(struct file *f, loff_t pos, size_t count);
1082 /* llite/lcommon_misc.c */
1083 int cl_ocd_update(struct obd_device *host, struct obd_device *watched,
1084 enum obd_notify_event ev, void *owner);
1085 int cl_get_grouplock(struct cl_object *obj, unsigned long gid, int nonblock,
1086 struct ll_grouplock *lg);
1087 void cl_put_grouplock(struct ll_grouplock *lg);
1089 /* llite/lproc_llite.c */
1090 int ll_debugfs_register_super(struct super_block *sb, const char *name);
1091 void ll_debugfs_unregister_super(struct super_block *sb);
1092 void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, long count);
1093 void ll_free_rw_stats_info(struct ll_sb_info *sbi);
1096 LPROC_LL_READ_BYTES,
1097 LPROC_LL_WRITE_BYTES,
1124 LPROC_LL_GETXATTR_HITS,
1126 LPROC_LL_REMOVEXATTR,
1127 LPROC_LL_INODE_PERM,
1129 LPROC_LL_INODE_OCOUNT,
1130 LPROC_LL_INODE_OPCLTM,
1131 LPROC_LL_FILE_OPCODES
1135 enum get_default_layout_type {
1136 GET_DEFAULT_LAYOUT_ROOT = 1,
1139 extern const struct file_operations ll_dir_operations;
1140 extern const struct inode_operations ll_dir_inode_operations;
1141 #ifdef HAVE_DIR_CONTEXT
1142 int ll_dir_read(struct inode *inode, __u64 *pos, struct md_op_data *op_data,
1143 struct dir_context *ctx, int *partial_readdir_rc);
1145 int ll_dir_read(struct inode *inode, __u64 *pos, struct md_op_data *op_data,
1146 void *cookie, filldir_t filldir, int *partial_readdir_rc);
1148 int ll_get_mdt_idx(struct inode *inode);
1149 int ll_get_mdt_idx_by_fid(struct ll_sb_info *sbi, const struct lu_fid *fid);
1150 struct page *ll_get_dir_page(struct inode *dir, struct md_op_data *op_data,
1151 __u64 offset, int *partial_readdir_rc);
1152 void ll_release_page(struct inode *inode, struct page *page, bool remove);
1153 int quotactl_ioctl(struct super_block *sb, struct if_quotactl *qctl);
1156 extern const struct inode_operations ll_special_inode_operations;
1158 struct inode *ll_iget(struct super_block *sb, ino_t hash,
1159 struct lustre_md *lic);
1160 int ll_test_inode_by_fid(struct inode *inode, void *opaque);
1161 int ll_md_blocking_ast(struct ldlm_lock *, struct ldlm_lock_desc *,
1162 void *data, int flag);
1163 struct dentry *ll_splice_alias(struct inode *inode, struct dentry *de);
1164 int ll_rmdir_entry(struct inode *dir, char *name, int namelen);
1165 void ll_update_times(struct ptlrpc_request *request, struct inode *inode);
1168 int ll_writepage(struct page *page, struct writeback_control *wbc);
1169 int ll_writepages(struct address_space *, struct writeback_control *wbc);
1170 int ll_readpage(struct file *file, struct page *page);
1171 #ifdef HAVE_AOPS_READ_FOLIO
1172 int ll_read_folio(struct file *file, struct folio *folio);
1174 int ll_io_read_page(const struct lu_env *env, struct cl_io *io,
1175 struct cl_page *page, struct file *file);
1176 void ll_readahead_init(struct inode *inode, struct ll_readahead_state *ras);
1177 int vvp_io_write_commit(const struct lu_env *env, struct cl_io *io);
1180 void ll_cl_add(struct inode *inode, const struct lu_env *env, struct cl_io *io,
1181 enum lcc_type type);
1182 void ll_cl_remove(struct inode *inode, const struct lu_env *env);
1183 struct ll_cl_context *ll_cl_find(struct inode *inode);
1185 extern const struct address_space_operations ll_aops;
1188 extern const struct inode_operations ll_file_inode_operations;
1189 const struct file_operations *ll_select_file_operations(struct ll_sb_info *sbi);
1190 extern int ll_have_md_lock(struct obd_export *exp, struct inode *inode,
1191 __u64 *bits, enum ldlm_mode l_req_mode);
1192 extern enum ldlm_mode ll_take_md_lock(struct inode *inode, __u64 bits,
1193 struct lustre_handle *lockh, __u64 flags,
1194 enum ldlm_mode mode);
1196 int ll_file_open(struct inode *inode, struct file *file);
1197 int ll_file_release(struct inode *inode, struct file *file);
1198 int ll_release_openhandle(struct dentry *, struct lookup_intent *);
1199 int ll_md_real_close(struct inode *inode, fmode_t fmode);
1200 void ll_track_file_opens(struct inode *inode);
1201 extern void ll_rw_stats_tally(struct ll_sb_info *sbi, pid_t pid,
1202 struct ll_file_data *file, loff_t pos,
1203 size_t count, int rw);
1204 #if defined(HAVE_USER_NAMESPACE_ARG) || defined(HAVE_INODEOPS_ENHANCED_GETATTR)
1205 int ll_getattr(struct user_namespace *mnt_userns, const struct path *path,
1206 struct kstat *stat, u32 request_mask, unsigned int flags);
1208 int ll_getattr(struct vfsmount *mnt, struct dentry *de, struct kstat *stat);
1209 #endif /* HAVE_USER_NAMESPACE_ARG */
1210 int ll_getattr_dentry(struct dentry *de, struct kstat *stat, u32 request_mask,
1211 unsigned int flags, bool foreign);
1212 #ifdef CONFIG_LUSTRE_FS_POSIX_ACL
1213 struct posix_acl *ll_get_acl(struct inode *inode, int type
1214 #ifdef HAVE_GET_ACL_RCU_ARG
1216 #endif /* HAVE_GET_ACL_RCU_ARG */
1218 int ll_set_acl(struct user_namespace *mnt_userns, struct inode *inode,
1219 struct posix_acl *acl, int type);
1220 #else /* !CONFIG_LUSTRE_FS_POSIX_ACL */
1221 #define ll_get_acl NULL
1222 #define ll_set_acl NULL
1223 #endif /* CONFIG_LUSTRE_FS_POSIX_ACL */
1225 static inline int ll_xflags_to_inode_flags(int xflags)
1227 return ((xflags & FS_XFLAG_SYNC) ? S_SYNC : 0) |
1228 ((xflags & FS_XFLAG_NOATIME) ? S_NOATIME : 0) |
1229 ((xflags & FS_XFLAG_APPEND) ? S_APPEND : 0) |
1230 ((xflags & FS_XFLAG_IMMUTABLE) ? S_IMMUTABLE : 0);
1233 static inline int ll_inode_flags_to_xflags(int inode_flags)
1235 return ((inode_flags & S_SYNC) ? FS_XFLAG_SYNC : 0) |
1236 ((inode_flags & S_NOATIME) ? FS_XFLAG_NOATIME : 0) |
1237 ((inode_flags & S_APPEND) ? FS_XFLAG_APPEND : 0) |
1238 ((inode_flags & S_IMMUTABLE) ? FS_XFLAG_IMMUTABLE : 0);
1241 int ll_migrate(struct inode *parent, struct file *file,
1242 struct lmv_user_md *lum, const char *name, __u32 flags);
1243 int ll_get_fid_by_name(struct inode *parent, const char *name,
1244 int namelen, struct lu_fid *fid, struct inode **inode);
1245 int ll_inode_permission(struct user_namespace *mnt_userns, struct inode *inode,
1247 int ll_ioctl_check_project(struct inode *inode, __u32 xflags, __u32 projid);
1248 int ll_ioctl_fsgetxattr(struct inode *inode, unsigned int cmd,
1250 int ll_ioctl_fssetxattr(struct inode *inode, unsigned int cmd,
1252 int ll_ioctl_project(struct file *file, unsigned int cmd, void __user *uarg);
1253 int ll_lov_setstripe_ea_info(struct inode *inode, struct dentry *dentry,
1254 __u64 flags, struct lov_user_md *lum,
1256 int ll_lov_getstripe_ea_info(struct inode *inode, const char *filename,
1257 struct lov_mds_md **lmm, int *lmm_size,
1258 struct ptlrpc_request **request);
1259 int ll_dir_setstripe(struct inode *inode, struct lov_user_md *lump,
1261 int ll_dir_get_default_layout(struct inode *inode, void **plmm, int *plmm_size,
1262 struct ptlrpc_request **request, u64 valid,
1263 enum get_default_layout_type type);
1264 int ll_dir_getstripe_default(struct inode *inode, void **lmmp,
1265 int *lmm_size, struct ptlrpc_request **request,
1266 struct ptlrpc_request **root_request, u64 valid);
1267 int ll_dir_getstripe(struct inode *inode, void **plmm, int *plmm_size,
1268 struct ptlrpc_request **request, u64 valid);
1269 int ll_fsync(struct file *file, loff_t start, loff_t end, int data);
1270 int ll_merge_attr(const struct lu_env *env, struct inode *inode);
1271 int ll_merge_attr_try(const struct lu_env *env, struct inode *inode);
1272 int ll_fid2path(struct inode *inode, void __user *arg);
1273 int __ll_fid2path(struct inode *inode, struct getinfo_fid2path *gfout,
1274 size_t outsize, __u32 pathlen_orig);
1275 int ll_data_version(struct inode *inode, __u64 *data_version, int flags);
1276 int ll_hsm_release(struct inode *inode);
1277 int ll_hsm_state_set(struct inode *inode, struct hsm_state_set *hss);
1278 void ll_io_set_mirror(struct cl_io *io, const struct file *file);
1280 /* llite/dcache.c */
1282 extern const struct dentry_operations ll_d_ops;
1284 bool ll_d_setup(struct dentry *de, bool do_put);
1286 static inline bool lld_is_init(struct dentry *dentry)
1288 return ll_d2d(dentry);
1291 #define ll_d_setup(de, do_put) (true)
1292 #define lld_is_init(dentry) (true)
1295 void ll_intent_drop_lock(struct lookup_intent *);
1296 void ll_intent_release(struct lookup_intent *);
1297 void ll_prune_aliases(struct inode *inode);
1298 void ll_lookup_finish_locks(struct lookup_intent *it, struct dentry *dentry);
1299 int ll_revalidate_it_finish(struct ptlrpc_request *request,
1300 struct lookup_intent *it, struct dentry *de);
1302 /* llite/llite_lib.c */
1303 extern const struct super_operations lustre_super_operations;
1305 void ll_lli_init(struct ll_inode_info *lli);
1306 int ll_fill_super(struct super_block *sb);
1307 void ll_put_super(struct super_block *sb);
1308 void ll_kill_super(struct super_block *sb);
1309 struct inode *ll_inode_from_resource_lock(struct ldlm_lock *lock);
1310 void ll_dir_clear_lsm_md(struct inode *inode);
1311 void ll_clear_inode(struct inode *inode);
1312 int volatile_ref_file(const char *volatile_name, int volatile_len,
1313 struct file **ref_file);
1314 int ll_setattr_raw(struct dentry *dentry, struct iattr *attr,
1315 enum op_xvalid xvalid, bool hsm_import);
1316 int ll_setattr(struct user_namespace *mnt_userns, struct dentry *de,
1317 struct iattr *attr);
1318 int ll_statfs(struct dentry *de, struct kstatfs *sfs);
1319 int ll_statfs_internal(struct ll_sb_info *sbi, struct obd_statfs *osfs,
1321 int ll_update_inode(struct inode *inode, struct lustre_md *md);
1322 void ll_update_inode_flags(struct inode *inode, unsigned int ext_flags);
1323 void ll_update_dir_depth_dmv(struct inode *dir, struct dentry *de);
1324 int ll_read_inode2(struct inode *inode, void *opaque);
1325 void ll_truncate_inode_pages_final(struct inode *inode, struct cl_io *io);
1326 void ll_delete_inode(struct inode *inode);
1327 int ll_iocontrol(struct inode *inode, struct file *file,
1328 unsigned int cmd, void __user *uarg);
1329 int ll_flush_ctx(struct inode *inode);
1330 void ll_umount_begin(struct super_block *sb);
1331 int ll_remount_fs(struct super_block *sb, int *flags, char *data);
1332 int ll_show_options(struct seq_file *seq, struct dentry *dentry);
1333 void ll_dirty_page_discard_warn(struct inode *inode, int ioret);
1334 int ll_prep_inode(struct inode **inode, struct req_capsule *pill,
1335 struct super_block *sb, struct lookup_intent *it);
1336 int ll_obd_statfs(struct inode *inode, void __user *arg);
1337 int ll_get_max_mdsize(struct ll_sb_info *sbi, int *max_mdsize);
1338 int ll_get_default_mdsize(struct ll_sb_info *sbi, int *default_mdsize);
1339 int ll_set_default_mdsize(struct ll_sb_info *sbi, int default_mdsize);
1341 void ll_unlock_md_op_lsm(struct md_op_data *op_data);
1342 struct md_op_data *ll_prep_md_op_data(struct md_op_data *op_data,
1343 struct inode *i1, struct inode *i2,
1344 const char *name, size_t namelen,
1345 __u32 mode, enum md_op_code opc,
1347 void ll_finish_md_op_data(struct md_op_data *op_data);
1348 int ll_get_obd_name(struct inode *inode, unsigned int cmd, void __user *arg);
1349 void ll_compute_rootsquash_state(struct ll_sb_info *sbi);
1350 ssize_t ll_copy_user_md(const struct lov_user_md __user *md,
1351 struct lov_user_md **kbuf);
1352 void ll_open_cleanup(struct super_block *sb, struct req_capsule *pill);
1354 void ll_dom_finish_open(struct inode *inode, struct ptlrpc_request *req);
1356 /* Compute expected user md size when passing in a md from user space */
1357 static inline ssize_t ll_lov_user_md_size(const struct lov_user_md *lum)
1359 switch (lum->lmm_magic) {
1360 case LOV_USER_MAGIC_V1:
1361 return sizeof(struct lov_user_md_v1);
1362 case LOV_USER_MAGIC_V3:
1363 return sizeof(struct lov_user_md_v3);
1364 case LOV_USER_MAGIC_SPECIFIC:
1365 if (lum->lmm_stripe_count > LOV_MAX_STRIPE_COUNT)
1368 return lov_user_md_size(lum->lmm_stripe_count,
1369 LOV_USER_MAGIC_SPECIFIC);
1370 case LOV_USER_MAGIC_COMP_V1:
1371 return ((struct lov_comp_md_v1 *)lum)->lcm_size;
1372 case LOV_USER_MAGIC_FOREIGN:
1373 return foreign_size(lum);
1379 /* llite/llite_nfs.c */
1380 extern const struct export_operations lustre_export_operations;
1381 __u32 get_uuid2int(const char *name, int len);
1382 struct inode *search_inode_for_lustre(struct super_block *sb,
1383 const struct lu_fid *fid);
1384 int ll_dir_get_parent_fid(struct inode *dir, struct lu_fid *parent_fid);
1386 /* llite/symlink.c */
1387 extern const struct inode_operations ll_fast_symlink_inode_operations;
1390 * IO arguments for various VFS I/O interfaces.
1392 struct vvp_io_args {
1393 /** normal/sendfile/splice */
1396 struct kiocb *via_iocb;
1397 struct iov_iter *via_iter;
1407 struct ll_cl_context {
1408 struct list_head lcc_list;
1410 const struct lu_env *lcc_env;
1411 struct cl_io *lcc_io;
1412 struct cl_page *lcc_page;
1413 enum lcc_type lcc_type;
1414 pgoff_t lcc_end_index;
1417 struct ll_thread_info {
1418 struct vvp_io_args lti_args;
1419 struct ra_io_arg lti_ria;
1420 struct ll_cl_context lti_io_ctx;
1423 extern struct lu_context_key ll_thread_key;
1425 static inline struct ll_thread_info *ll_env_info(const struct lu_env *env)
1427 struct ll_thread_info *lti;
1429 lti = lu_context_key_get(&env->le_ctx, &ll_thread_key);
1430 LASSERT(lti != NULL);
1435 static inline struct vvp_io_args *ll_env_args(const struct lu_env *env)
1437 return &ll_env_info(env)->lti_args;
1440 void ll_io_init(struct cl_io *io, struct file *file, enum cl_io_type iot,
1441 struct vvp_io_args *args);
1443 /* llite/llite_mmap.c */
1445 int ll_file_mmap(struct file * file, struct vm_area_struct * vma);
1446 void policy_from_vma(union ldlm_policy_data *policy, struct vm_area_struct *vma,
1447 unsigned long addr, size_t count);
1448 struct vm_area_struct *our_vma(struct mm_struct *mm, unsigned long addr,
1451 #define ll_s2sbi(sb) (s2lsi(sb)->lsi_llsbi)
1453 /* don't need an addref as the sb_info should be holding one */
1454 static inline struct obd_export *ll_s2dtexp(struct super_block *sb)
1456 return ll_s2sbi(sb)->ll_dt_exp;
1459 /* don't need an addref as the sb_info should be holding one */
1460 static inline struct obd_export *ll_s2mdexp(struct super_block *sb)
1462 return ll_s2sbi(sb)->ll_md_exp;
1465 static inline struct client_obd *sbi2mdc(struct ll_sb_info *sbi)
1467 struct obd_device *obd = sbi->ll_md_exp->exp_obd;
1473 // FIXME: replace the name of this with LL_SB to conform to kernel stuff
1474 static inline struct ll_sb_info *ll_i2sbi(struct inode *inode)
1476 return ll_s2sbi(inode->i_sb);
1479 static inline struct obd_export *ll_i2dtexp(struct inode *inode)
1481 return ll_s2dtexp(inode->i_sb);
1484 static inline struct obd_export *ll_i2mdexp(struct inode *inode)
1486 return ll_s2mdexp(inode->i_sb);
1489 static inline struct lu_fid *ll_inode2fid(struct inode *inode)
1493 LASSERT(inode != NULL);
1494 fid = &ll_i2info(inode)->lli_fid;
1499 static inline bool ll_dir_striped(struct inode *inode)
1501 struct ll_inode_info *lli;
1505 if (!S_ISDIR(inode->i_mode))
1508 lli = ll_i2info(inode);
1509 if (!lli->lli_lsm_md)
1512 down_read(&lli->lli_lsm_sem);
1513 rc = lmv_dir_striped(lli->lli_lsm_md);
1514 up_read(&lli->lli_lsm_sem);
1519 static inline loff_t ll_file_maxbytes(struct inode *inode)
1521 struct cl_object *obj = ll_i2info(inode)->lli_clob;
1524 return MAX_LFS_FILESIZE;
1526 return min_t(loff_t, cl_object_maxbytes(obj), MAX_LFS_FILESIZE);
1530 extern const struct xattr_handler *ll_xattr_handlers[];
1532 #define XATTR_USER_T 1
1533 #define XATTR_TRUSTED_T 2
1534 #define XATTR_SECURITY_T 3
1535 #define XATTR_ACL_ACCESS_T 4
1536 #define XATTR_ACL_DEFAULT_T 5
1537 #define XATTR_LUSTRE_T 6
1538 #define XATTR_OTHER_T 7
1539 #define XATTR_ENCRYPTION_T 9
1541 ssize_t ll_listxattr(struct dentry *dentry, char *buffer, size_t size);
1542 int ll_xattr_list(struct inode *inode, const char *name, int type,
1543 void *buffer, size_t size, u64 valid);
1544 const struct xattr_handler *get_xattr_type(const char *name);
1547 * Common IO arguments for various VFS I/O interfaces.
1549 int cl_sb_init(struct super_block *sb);
1550 int cl_sb_fini(struct super_block *sb);
1552 enum ras_update_flags {
1556 void ll_ra_count_put(struct ll_sb_info *sbi, unsigned long len);
1557 void ll_ra_stats_inc(struct inode *inode, enum ra_stat which);
1561 #define LL_SA_RPC_MIN 8
1562 #define LL_SA_RPC_DEF 128
1563 #define LL_SA_RPC_MAX 2048
1565 /* XXX: If want to support more concurrent statahead instances,
1566 * please consider to decentralize the RPC lists attached
1567 * on related import, such as imp_{sending,delayed}_list.
1569 #define LL_SA_RUNNING_MAX 256
1570 #define LL_SA_RUNNING_DEF 16
1572 #define LL_SA_BATCH_MAX 1024
1573 #define LL_SA_BATCH_DEF 64
1575 #define LL_SA_CACHE_BIT 5
1576 #define LL_SA_CACHE_SIZE (1 << LL_SA_CACHE_BIT)
1577 #define LL_SA_CACHE_MASK (LL_SA_CACHE_SIZE - 1)
1579 /* per inode struct, for dir only */
1580 struct ll_statahead_info {
1581 struct dentry *sai_dentry;
1582 atomic_t sai_refcount; /* when access this struct, hold
1584 unsigned int sai_max; /* max ahead of lookup */
1585 __u64 sai_sent; /* stat requests sent count */
1586 __u64 sai_replied; /* stat requests which received
1588 __u64 sai_index; /* index of statahead entry */
1589 __u64 sai_index_wait; /* index of entry which is the
1590 * caller is waiting for */
1591 __u64 sai_hit; /* hit count */
1592 __u64 sai_miss; /* miss count:
1593 * for "ls -al" case, includes
1594 * hidden dentry miss;
1595 * for "ls -l" case, it does not
1596 * include hidden dentry miss.
1597 * "sai_miss_hidden" is used for
1600 unsigned int sai_consecutive_miss; /* consecutive miss */
1601 unsigned int sai_miss_hidden;/* "ls -al", but first dentry
1602 * is not a hidden one */
1603 unsigned int sai_skip_hidden;/* skipped hidden dentry count
1605 unsigned int sai_ls_all:1; /* "ls -al", do stat-ahead for
1607 wait_queue_head_t sai_waitq; /* stat-ahead wait queue */
1608 struct task_struct *sai_task; /* stat-ahead thread */
1609 struct task_struct *sai_agl_task; /* AGL thread */
1610 struct list_head sai_entries; /* completed entries */
1611 struct list_head sai_agls; /* AGLs to be sent */
1612 struct list_head sai_cache[LL_SA_CACHE_SIZE];
1613 spinlock_t sai_cache_lock[LL_SA_CACHE_SIZE];
1614 atomic_t sai_cache_count; /* entry count in cache */
1615 struct lu_batch *sai_bh;
1616 __u32 sai_max_batch_count;
1617 __u64 sai_index_end;
1620 int ll_revalidate_statahead(struct inode *dir, struct dentry **dentry,
1622 int ll_start_statahead(struct inode *dir, struct dentry *dentry, bool agl);
1623 void ll_authorize_statahead(struct inode *dir, void *key);
1624 void ll_deauthorize_statahead(struct inode *dir, void *key);
1627 blkcnt_t dirty_cnt(struct inode *inode);
1629 int cl_glimpse_size0(struct inode *inode, int agl);
1630 int cl_glimpse_lock(const struct lu_env *env, struct cl_io *io,
1631 struct inode *inode, struct cl_object *clob, int agl);
1633 static inline int cl_glimpse_size(struct inode *inode)
1635 return cl_glimpse_size0(inode, 0);
1638 /* AGL is 'asychronous glimpse lock', which is a speculative lock taken as
1639 * part of statahead */
1640 static inline int cl_agl(struct inode *inode)
1642 return cl_glimpse_size0(inode, 1);
1645 int ll_file_lock_ahead(struct file *file, struct llapi_lu_ladvise *ladvise);
1647 int cl_io_get(struct inode *inode, struct lu_env **envout,
1648 struct cl_io **ioout, __u16 *refcheck);
1650 static inline int ll_glimpse_size(struct inode *inode)
1652 struct ll_inode_info *lli = ll_i2info(inode);
1655 down_read(&lli->lli_glimpse_sem);
1656 rc = cl_glimpse_size(inode);
1657 lli->lli_glimpse_time = ktime_get();
1658 up_read(&lli->lli_glimpse_sem);
1662 /* dentry may statahead when statahead is enabled and current process has opened
1663 * parent directory, and this dentry hasn't accessed statahead cache before */
1665 dentry_may_statahead(struct inode *dir, struct dentry *dentry)
1667 struct ll_inode_info *lli;
1668 struct ll_dentry_data *ldd;
1670 if (ll_i2sbi(dir)->ll_sa_max == 0)
1673 lli = ll_i2info(dir);
1675 /* statahead is not allowed for this dir, there may be three causes:
1676 * 1. dir is not opened.
1677 * 2. statahead hit ratio is too low.
1678 * 3. previous stat started statahead thread failed. */
1679 if (!lli->lli_sa_enabled)
1682 /* not the same process, don't statahead */
1683 if (lli->lli_opendir_pid != current->pid)
1687 * When stating a dentry, kernel may trigger 'revalidate' or 'lookup'
1688 * multiple times, eg. for 'getattr', 'getxattr' and etc.
1689 * For patchless client, lookup intent is not accurate, which may
1690 * misguide statahead. For example:
1691 * The 'revalidate' call for 'getattr' and 'getxattr' of a dentry will
1692 * have the same intent -- IT_GETATTR, while one dentry should access
1693 * statahead cache once, otherwise statahead windows is messed up.
1694 * The solution is as following:
1695 * Assign 'lld_sa_generation' with 'lli_sa_generation' when a dentry
1696 * IT_GETATTR for the first time, and subsequent IT_GETATTR will
1697 * bypass interacting with statahead cache by checking
1698 * 'lld_sa_generation == lli->lli_sa_generation'.
1700 ldd = ll_d2d(dentry);
1701 if (ldd != NULL && lli->lli_sa_generation &&
1702 ldd->lld_sa_generation == lli->lli_sa_generation)
1708 int cl_sync_file_range(struct inode *inode, loff_t start, loff_t end,
1709 enum cl_fsync_mode mode, int ignore_layout);
1711 static inline int ll_file_nolock(const struct file *file)
1713 struct ll_file_data *fd = file->private_data;
1714 struct inode *inode = file_inode((struct file *)file);
1716 LASSERT(fd != NULL);
1717 return ((fd->fd_flags & LL_FILE_IGNORE_LOCK) ||
1718 test_bit(LL_SBI_NOLCK, ll_i2sbi(inode)->ll_flags));
1721 static inline void ll_set_lock_data(struct obd_export *exp, struct inode *inode,
1722 struct lookup_intent *it, __u64 *bits)
1724 if (!it->it_lock_set) {
1725 struct lustre_handle handle;
1727 /* If this inode is a remote object, it will get two
1728 * separate locks in different namespaces, Master MDT,
1729 * where the name entry is, will grant LOOKUP lock,
1730 * remote MDT, where the object is, will grant
1731 * UPDATE|PERM lock. The inode will be attched to both
1732 * LOOKUP and PERM locks, so revoking either locks will
1733 * case the dcache being cleared */
1734 if (it->it_remote_lock_mode) {
1735 handle.cookie = it->it_remote_lock_handle;
1736 CDEBUG(D_DLMTRACE, "setting l_data to inode "DFID
1737 "(%p) for remote lock %#llx\n",
1738 PFID(ll_inode2fid(inode)), inode,
1740 md_set_lock_data(exp, &handle, inode, NULL);
1743 handle.cookie = it->it_lock_handle;
1745 CDEBUG(D_DLMTRACE, "setting l_data to inode "DFID"(%p)"
1746 " for lock %#llx\n",
1747 PFID(ll_inode2fid(inode)), inode, handle.cookie);
1749 md_set_lock_data(exp, &handle, inode, &it->it_lock_bits);
1750 it->it_lock_set = 1;
1754 *bits = it->it_lock_bits;
1757 static inline int d_lustre_invalid(const struct dentry *dentry)
1759 return !ll_d2d(dentry) || ll_d2d(dentry)->lld_invalid;
1763 * Mark dentry INVALID, if dentry refcount is zero (this is normally case for
1764 * ll_md_blocking_ast), it will be pruned by ll_prune_aliases() and
1765 * ll_prune_negative_children(); otherwise dput() of the last refcount will
1766 * unhash this dentry and kill it.
1768 static inline void d_lustre_invalidate(struct dentry *dentry)
1770 CDEBUG(D_DENTRY, "invalidate dentry %pd (%p) parent %p inode %p refc %d\n",
1772 dentry->d_parent, dentry->d_inode, ll_d_count(dentry));
1774 spin_lock(&dentry->d_lock);
1775 if (lld_is_init(dentry))
1776 ll_d2d(dentry)->lld_invalid = 1;
1777 spin_unlock(&dentry->d_lock);
1780 static inline void d_lustre_revalidate(struct dentry *dentry)
1782 spin_lock(&dentry->d_lock);
1783 LASSERT(ll_d2d(dentry));
1784 ll_d2d(dentry)->lld_invalid = 0;
1785 spin_unlock(&dentry->d_lock);
1788 static inline dev_t ll_compat_encode_dev(dev_t dev)
1790 /* The compat_sys_*stat*() syscalls will fail unless the
1791 * device majors and minors are both less than 256. Note that
1792 * the value returned here will be passed through
1793 * old_encode_dev() in cp_compat_stat(). And so we are not
1794 * trying to return a valid compat (u16) device number, just
1795 * one that will pass the old_valid_dev() check. */
1797 return MKDEV(MAJOR(dev) & 0xff, MINOR(dev) & 0xff);
1800 int ll_layout_conf(struct inode *inode, const struct cl_object_conf *conf);
1801 int ll_layout_refresh(struct inode *inode, __u32 *gen);
1802 int ll_layout_restore(struct inode *inode, loff_t start, __u64 length);
1803 int ll_layout_write_intent(struct inode *inode, enum layout_intent_opc opc,
1804 struct lu_extent *ext);
1806 int ll_xattr_init(void);
1807 void ll_xattr_fini(void);
1809 int ll_page_sync_io(const struct lu_env *env, struct cl_io *io,
1810 struct cl_page *page, enum cl_req_type crt);
1812 int ll_getparent(struct file *file, struct getparent __user *arg);
1815 int cl_setattr_ost(struct cl_object *obj, const struct iattr *attr,
1816 enum op_xvalid xvalid, unsigned int attr_flags);
1818 extern struct lu_env *cl_inode_fini_env;
1819 extern __u16 cl_inode_fini_refcheck;
1821 int cl_file_inode_init(struct inode *inode, struct lustre_md *md);
1822 void cl_inode_fini(struct inode *inode);
1824 u64 cl_fid_build_ino(const struct lu_fid *fid, int api32);
1825 u32 cl_fid_build_gen(const struct lu_fid *fid);
1827 static inline struct pcc_super *ll_i2pccs(struct inode *inode)
1829 return &ll_i2sbi(inode)->ll_pcc_super;
1832 static inline struct pcc_super *ll_info2pccs(struct ll_inode_info *lli)
1834 return ll_i2pccs(ll_info2i(lli));
1838 /* The digested form is made of a FID (16 bytes) followed by the second-to-last
1839 * ciphertext block (16 bytes), so a total length of 32 bytes.
1840 * That way, llcrypt does not compute a digested form of this digest.
1842 struct ll_digest_filename {
1843 struct lu_fid ldf_fid;
1844 char ldf_excerpt[LL_CRYPTO_BLOCK_SIZE];
1847 int ll_prepare_lookup(struct inode *dir, struct dentry *de,
1848 struct llcrypt_name *fname, struct lu_fid *fid);
1849 int ll_setup_filename(struct inode *dir, const struct qstr *iname,
1850 int lookup, struct llcrypt_name *fname,
1851 struct lu_fid *fid);
1852 int ll_fname_disk_to_usr(struct inode *inode,
1853 u32 hash, u32 minor_hash,
1854 struct llcrypt_str *iname, struct llcrypt_str *oname,
1855 struct lu_fid *fid);
1856 int ll_file_open_encrypt(struct inode *inode, struct file *filp);
1858 static inline char *xattr_for_enc(struct inode *inode)
1860 if (ll_sbi_has_name_encrypt(ll_i2sbi(inode)))
1861 return LL_XATTR_NAME_ENCRYPTION_CONTEXT;
1863 return LL_XATTR_NAME_ENCRYPTION_CONTEXT_OLD;
1866 #ifdef HAVE_LUSTRE_CRYPTO
1867 extern const struct llcrypt_operations lustre_cryptops;
1870 /* llite/llite_foreign.c */
1871 int ll_manage_foreign(struct inode *inode, struct lustre_md *lmd);
1872 bool ll_foreign_is_openable(struct dentry *dentry, unsigned int flags);
1873 bool ll_foreign_is_removable(struct dentry *dentry, bool unset);
1875 int ll_filemap_fault(struct vm_area_struct *vma, struct vm_fault *vmf);
1877 #endif /* LLITE_INTERNAL_H */