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 <lustre_compat.h>
46 #include <lustre_crypto.h>
47 #include <range_lock.h>
49 #include "vvp_internal.h"
51 #include "foreign_symlink.h"
57 #ifndef HAVE_VM_FAULT_RETRY
58 #define VM_FAULT_RETRY 0
61 /* Kernel 3.1 kills LOOKUP_CONTINUE, LOOKUP_PARENT is equivalent to it.
62 * seem kernel commit 49084c3bb2055c401f3493c13edae14d49128ca0 */
63 #ifndef LOOKUP_CONTINUE
64 #define LOOKUP_CONTINUE LOOKUP_PARENT
67 /** Only used on client-side for indicating the tail of dir hash/offset. */
68 #define LL_DIR_END_OFF 0x7fffffffffffffffULL
69 #define LL_DIR_END_OFF_32BIT 0x7fffffffUL
71 /* 4UL * 1024 * 1024 */
72 #define LL_MAX_BLKSIZE_BITS 22
74 #define LL_IT2STR(it) ((it) ? ldlm_it2str((it)->it_op) : "0")
76 #define TIMES_SET_FLAGS (ATTR_MTIME_SET | ATTR_ATIME_SET | ATTR_TIMES_SET)
78 struct ll_dentry_data {
79 unsigned int lld_sa_generation;
80 unsigned int lld_invalid:1;
81 unsigned int lld_nfs_dentry:1;
82 struct rcu_head lld_rcu_head;
85 #define ll_d2d(de) ((struct ll_dentry_data*)((de)->d_fsdata))
87 #define LLI_INODE_MAGIC 0x111d0de5
88 #define LLI_INODE_DEAD 0xdeadd00d
90 struct ll_getname_data {
91 #ifdef HAVE_DIR_CONTEXT
92 struct dir_context ctx;
94 char *lgd_name; /* points to a buffer with NAME_MAX+1 size */
95 struct lu_fid lgd_fid; /* target fid we are looking for */
96 int lgd_found; /* inode matched? */
100 struct lu_env *lg_env;
102 struct cl_lock *lg_lock;
103 unsigned long lg_gid;
106 /* See comment on trunc_sem_down_read_nowait */
107 struct ll_trunc_sem {
108 /* when positive, this is a count of readers, when -1, it indicates
109 * the semaphore is held for write, and 0 is unlocked
111 atomic_t ll_trunc_readers;
112 /* this tracks a count of waiting writers */
113 atomic_t ll_trunc_waiters;
116 struct ll_inode_info {
117 __u32 lli_inode_magic;
120 volatile unsigned long lli_flags;
121 struct posix_acl *lli_posix_acl;
123 /* identifying fields for both metadata and data stacks. */
124 struct lu_fid lli_fid;
125 /* master inode fid for stripe directory */
126 struct lu_fid lli_pfid;
128 /* We need all three because every inode may be opened in different
130 struct obd_client_handle *lli_mds_read_och;
131 struct obd_client_handle *lli_mds_write_och;
132 struct obd_client_handle *lli_mds_exec_och;
133 __u64 lli_open_fd_read_count;
134 __u64 lli_open_fd_write_count;
135 __u64 lli_open_fd_exec_count;
137 /* Number of times this inode was opened */
138 u64 lli_open_fd_count;
139 /* When last close was performed on this inode */
140 ktime_t lli_close_fd_time;
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;
154 /* Try to make the d::member and f::member are aligned. Before using
155 * these members, make clear whether it is directory or not. */
159 /* metadata statahead */
160 /* since parent-child threads can share the same @file
161 * struct, "opendir_key" is the token when dir close for
162 * case of parent exit before child -- it is me should
163 * cleanup the dir readahead. */
164 void *lli_opendir_key;
165 struct ll_statahead_info *lli_sai;
166 /* protect statahead stuff. */
167 spinlock_t lli_sa_lock;
168 /* "opendir_pid" is the token when lookup/revalid
169 * -- I am the owner of dir statahead. */
170 pid_t lli_opendir_pid;
171 /* stat will try to access statahead entries or start
172 * statahead if this flag is set, and this flag will be
173 * set upon dir open, and cleared when dir is closed,
174 * statahead hit ratio is too low, or start statahead
176 unsigned int lli_sa_enabled:1;
177 /* generation for statahead */
178 unsigned int lli_sa_generation;
179 /* rw lock protects lli_lsm_md */
180 struct rw_semaphore lli_lsm_sem;
181 /* directory stripe information */
182 struct lmv_stripe_md *lli_lsm_md;
183 /* directory default LMV */
184 struct lmv_stripe_md *lli_default_lsm_md;
187 /* for non-directory */
189 struct mutex lli_size_mutex;
190 char *lli_symlink_name;
191 struct ll_trunc_sem lli_trunc_sem;
192 struct range_lock_tree lli_write_tree;
193 struct mutex lli_setattr_mutex;
195 struct rw_semaphore lli_glimpse_sem;
196 ktime_t lli_glimpse_time;
197 struct list_head lli_agl_list;
200 /* for writepage() only to communicate to fsync */
203 /* protect the file heat fields */
204 spinlock_t lli_heat_lock;
205 __u32 lli_heat_flags;
206 struct obd_heat_instance lli_heat_instances[OBD_HEAT_COUNT];
209 * Whenever a process try to read/write the file, the
210 * jobid of the process will be saved here, and it'll
211 * be packed into the write PRC when flush later.
213 * So the read/write statistics for jobid will not be
214 * accurate if the file is shared by different jobs.
216 char lli_jobid[LUSTRE_JOBID_SIZE];
218 struct mutex lli_pcc_lock;
219 enum lu_pcc_state_flags lli_pcc_state;
221 * @lli_pcc_generation saves the gobal PCC generation
222 * when the file was successfully attached into PCC.
223 * The flags of the PCC dataset are saved in
225 * The gobal PCC generation will be increased when add
226 * or delete a PCC backend, or change the configuration
227 * parameters for PCC.
228 * If @lli_pcc_generation is same as the gobal PCC
229 * generation, we can use the saved flags of the PCC
230 * dataset to determine whether need to try auto attach
233 __u64 lli_pcc_generation;
234 enum pcc_dataset_flags lli_pcc_dsflags;
235 struct pcc_inode *lli_pcc_inode;
237 struct mutex lli_group_mutex;
238 __u64 lli_group_users;
239 unsigned long lli_group_gid;
241 __u64 lli_attr_valid;
243 __u64 lli_lazyblocks;
247 /* XXX: For following frequent used members, although they maybe special
248 * used for non-directory object, it is some time-wasting to check
249 * whether the object is directory or not before using them. On the
250 * other hand, currently, sizeof(f) > sizeof(d), it cannot reduce
251 * the "ll_inode_info" size even if moving those members into u.f.
252 * So keep them out side.
254 * In the future, if more members are added only for directory,
255 * some of the following members can be moved into u.f.
257 struct cl_object *lli_clob;
259 /* mutex to request for layout lock exclusively. */
260 struct mutex lli_layout_mutex;
261 /* Layout version, protected by lli_layout_lock */
262 __u32 lli_layout_gen;
263 spinlock_t lli_layout_lock;
265 __u32 lli_projid; /* project id */
267 struct rw_semaphore lli_xattrs_list_rwsem;
268 struct mutex lli_xattrs_enq_lock;
269 struct list_head lli_xattrs; /* ll_xattr_entry->xe_list */
272 static inline void ll_trunc_sem_init(struct ll_trunc_sem *sem)
274 atomic_set(&sem->ll_trunc_readers, 0);
275 atomic_set(&sem->ll_trunc_waiters, 0);
278 /* This version of down read ignores waiting writers, meaning if the semaphore
279 * is already held for read, this down_read will 'join' that reader and also
280 * take the semaphore.
282 * This lets us avoid an unusual deadlock.
284 * We must take lli_trunc_sem in read mode on entry in to various i/o paths
285 * in Lustre, in order to exclude truncates. Some of these paths then need to
286 * take the mmap_lock, while still holding the trunc_sem. The problem is that
287 * page faults hold the mmap_lock when calling in to Lustre, and then must also
288 * take the trunc_sem to exclude truncate.
290 * This means the locking order for trunc_sem and mmap_lock is sometimes AB,
291 * sometimes BA. This is almost OK because in both cases, we take the trunc
292 * sem for read, so it doesn't block.
294 * However, if a write mode user (truncate, a setattr op) arrives in the
295 * middle of this, the second reader on the truncate_sem will wait behind that
298 * So we have, on our truncate sem, in order (where 'reader' and 'writer' refer
299 * to the mode in which they take the semaphore):
300 * reader (holding mmap_lock, needs truncate_sem)
302 * reader (holding truncate sem, waiting for mmap_lock)
304 * And so the readers deadlock.
306 * The solution is this modified semaphore, where this down_read ignores
307 * waiting write operations, and all waiters are woken up at once, so readers
308 * using down_read_nowait cannot get stuck behind waiting writers, regardless
309 * of the order they arrived in.
311 * down_read_nowait is only used in the page fault case, where we already hold
312 * the mmap_lock. This is because otherwise repeated read and write operations
313 * (which take the truncate sem) could prevent a truncate from ever starting.
314 * This could still happen with page faults, but without an even more complex
315 * mechanism, this is unavoidable.
319 static inline void trunc_sem_down_read_nowait(struct ll_trunc_sem *sem)
321 wait_var_event(&sem->ll_trunc_readers,
322 atomic_inc_unless_negative(&sem->ll_trunc_readers));
325 static inline void trunc_sem_down_read(struct ll_trunc_sem *sem)
327 wait_var_event(&sem->ll_trunc_readers,
328 atomic_read(&sem->ll_trunc_waiters) == 0 &&
329 atomic_inc_unless_negative(&sem->ll_trunc_readers));
332 static inline void trunc_sem_up_read(struct ll_trunc_sem *sem)
334 if (atomic_dec_return(&sem->ll_trunc_readers) == 0 &&
335 atomic_read(&sem->ll_trunc_waiters))
336 wake_up_var(&sem->ll_trunc_readers);
339 static inline void trunc_sem_down_write(struct ll_trunc_sem *sem)
341 atomic_inc(&sem->ll_trunc_waiters);
342 wait_var_event(&sem->ll_trunc_readers,
343 atomic_cmpxchg(&sem->ll_trunc_readers, 0, -1) == 0);
344 atomic_dec(&sem->ll_trunc_waiters);
347 static inline void trunc_sem_up_write(struct ll_trunc_sem *sem)
349 atomic_set(&sem->ll_trunc_readers, 0);
350 wake_up_var(&sem->ll_trunc_readers);
353 #ifdef CONFIG_LUSTRE_FS_POSIX_ACL
354 static inline void lli_clear_acl(struct ll_inode_info *lli)
356 if (lli->lli_posix_acl) {
357 posix_acl_release(lli->lli_posix_acl);
358 lli->lli_posix_acl = NULL;
362 static inline void lli_replace_acl(struct ll_inode_info *lli,
363 struct lustre_md *md)
365 spin_lock(&lli->lli_lock);
366 if (lli->lli_posix_acl)
367 posix_acl_release(lli->lli_posix_acl);
368 lli->lli_posix_acl = md->posix_acl;
369 spin_unlock(&lli->lli_lock);
372 static inline void lli_clear_acl(struct ll_inode_info *lli)
376 static inline void lli_replace_acl(struct ll_inode_info *lli,
377 struct lustre_md *md)
382 static inline __u32 ll_layout_version_get(struct ll_inode_info *lli)
386 spin_lock(&lli->lli_layout_lock);
387 gen = lli->lli_layout_gen;
388 spin_unlock(&lli->lli_layout_lock);
393 static inline void ll_layout_version_set(struct ll_inode_info *lli, __u32 gen)
395 spin_lock(&lli->lli_layout_lock);
396 lli->lli_layout_gen = gen;
397 spin_unlock(&lli->lli_layout_lock);
401 /* File data is modified. */
402 LLIF_DATA_MODIFIED = 0,
403 /* File is being restored */
404 LLIF_FILE_RESTORING = 1,
405 /* Xattr cache is attached to the file */
406 LLIF_XATTR_CACHE = 2,
407 /* Project inherit */
408 LLIF_PROJECT_INHERIT = 3,
409 /* update atime from MDS even if it's older than local inode atime. */
410 LLIF_UPDATE_ATIME = 4,
411 /* foreign file/dir can be unlinked unconditionnaly */
412 LLIF_FOREIGN_REMOVABLE = 5,
413 /* setting encryption context in progress */
414 LLIF_SET_ENC_CTX = 6,
418 int ll_xattr_cache_destroy(struct inode *inode);
420 int ll_xattr_cache_get(struct inode *inode,
426 int ll_xattr_cache_insert(struct inode *inode,
431 static inline bool obd_connect_has_secctx(struct obd_connect_data *data)
433 #ifdef CONFIG_SECURITY
434 return data->ocd_connect_flags & OBD_CONNECT_FLAGS2 &&
435 data->ocd_connect_flags2 & OBD_CONNECT2_FILE_SECCTX;
441 static inline void obd_connect_set_secctx(struct obd_connect_data *data)
443 #ifdef CONFIG_SECURITY
444 data->ocd_connect_flags2 |= OBD_CONNECT2_FILE_SECCTX;
448 int ll_dentry_init_security(struct dentry *dentry, int mode, struct qstr *name,
449 const char **secctx_name, void **secctx,
451 int ll_inode_init_security(struct dentry *dentry, struct inode *inode,
454 int ll_listsecurity(struct inode *inode, char *secctx_name,
455 size_t secctx_name_size);
457 static inline bool obd_connect_has_enc(struct obd_connect_data *data)
459 #ifdef HAVE_LUSTRE_CRYPTO
460 return data->ocd_connect_flags & OBD_CONNECT_FLAGS2 &&
461 data->ocd_connect_flags2 & OBD_CONNECT2_ENCRYPT;
467 static inline void obd_connect_set_enc(struct obd_connect_data *data)
469 #ifdef HAVE_LUSTRE_CRYPTO
470 data->ocd_connect_flags2 |= OBD_CONNECT2_ENCRYPT;
475 * Locking to guarantee consistency of non-atomic updates to long long i_size,
476 * consistency between file size and KMS.
478 * Implemented by ->lli_size_mutex and ->lsm_lock, nested in that order.
481 void ll_inode_size_lock(struct inode *inode);
482 void ll_inode_size_unlock(struct inode *inode);
484 static inline struct ll_inode_info *ll_i2info(struct inode *inode)
486 return container_of(inode, struct ll_inode_info, lli_vfs_inode);
489 static inline struct pcc_inode *ll_i2pcci(struct inode *inode)
491 return ll_i2info(inode)->lli_pcc_inode;
494 /* default to use at least 16M for fast read if possible */
495 #define RA_REMAIN_WINDOW_MIN MiB_TO_PAGES(16UL)
497 /* default read-ahead on a given client mountpoint. */
498 #define SBI_DEFAULT_READ_AHEAD_MAX MiB_TO_PAGES(1024UL)
500 /* default read-ahead for a single file descriptor */
501 #define SBI_DEFAULT_READ_AHEAD_PER_FILE_MAX MiB_TO_PAGES(256UL)
503 /* default read-ahead full files smaller than limit on the second read */
504 #define SBI_DEFAULT_READ_AHEAD_WHOLE_MAX MiB_TO_PAGES(2UL)
506 /* default range pages */
507 #define SBI_DEFAULT_RA_RANGE_PAGES MiB_TO_PAGES(1ULL)
509 /* Min range pages */
510 #define RA_MIN_MMAP_RANGE_PAGES 16UL
515 RA_STAT_DISTANT_READPAGE,
516 RA_STAT_MISS_IN_WINDOW,
517 RA_STAT_FAILED_GRAB_PAGE,
518 RA_STAT_FAILED_MATCH,
523 RA_STAT_MAX_IN_FLIGHT,
524 RA_STAT_WRONG_GRAB_PAGE,
525 RA_STAT_FAILED_REACH_END,
527 RA_STAT_FAILED_FAST_READ,
528 RA_STAT_MMAP_RANGE_READ,
533 atomic_t ra_cur_pages;
534 unsigned long ra_max_pages;
535 unsigned long ra_max_pages_per_file;
536 unsigned long ra_range_pages;
537 unsigned long ra_max_read_ahead_whole_pages;
538 struct workqueue_struct *ll_readahead_wq;
540 * Max number of active works could be triggered
541 * for async readahead.
543 unsigned int ra_async_max_active;
544 /* how many async readahead triggered in flight */
545 atomic_t ra_async_inflight;
546 /* Threshold to control when to trigger async readahead */
547 unsigned long ra_async_pages_per_file_threshold;
550 /* ra_io_arg will be filled in the beginning of ll_readahead with
551 * ras_lock, then the following ll_read_ahead_pages will read RA
552 * pages according to this arg, all the items in this structure are
553 * counted by page index.
556 pgoff_t ria_start_idx; /* start offset of read-ahead*/
557 pgoff_t ria_end_idx; /* end offset of read-ahead*/
558 unsigned long ria_reserved; /* reserved pages for read-ahead */
559 pgoff_t ria_end_idx_min;/* minimum end to cover current read */
560 bool ria_eof; /* reach end of file */
561 /* If stride read pattern is detected, ria_stoff is the byte offset
562 * where stride read is started. Note: for normal read-ahead, the
563 * value here is meaningless, and also it will not be accessed*/
565 /* ria_length and ria_bytes are the length and pages length in the
566 * stride I/O mode. And they will also be used to check whether
567 * it is stride I/O read-ahead in the read-ahead pages*/
572 /* LL_HIST_MAX=32 causes an overflow */
573 #define LL_HIST_MAX 28
574 #define LL_HIST_START 12 /* buckets start at 2^12 = 4k */
575 #define LL_PROCESS_HIST_MAX 10
576 struct per_process_info {
578 struct obd_histogram pp_r_hist;
579 struct obd_histogram pp_w_hist;
582 /* pp_extents[LL_PROCESS_HIST_MAX] will hold the combined process info */
583 struct ll_rw_extents_info {
584 struct per_process_info pp_extents[LL_PROCESS_HIST_MAX + 1];
587 #define LL_OFFSET_HIST_MAX 100
588 struct ll_rw_process_info {
591 loff_t rw_range_start;
593 loff_t rw_last_file_pos;
595 size_t rw_smallest_extent;
596 size_t rw_largest_extent;
597 struct ll_file_data *rw_last_file;
600 enum stats_track_type {
601 STATS_TRACK_ALL = 0, /* track all processes */
602 STATS_TRACK_PID, /* track process with this pid */
603 STATS_TRACK_PPID, /* track processes with this ppid */
604 STATS_TRACK_GID, /* track processes with this gid */
608 /* flags for sbi->ll_flags */
609 #define LL_SBI_NOLCK 0x01 /* DLM locking disabled (directio-only) */
610 #define LL_SBI_CHECKSUM 0x02 /* checksum each page as it's written */
611 #define LL_SBI_FLOCK 0x04
612 #define LL_SBI_USER_XATTR 0x08 /* support user xattr */
613 #define LL_SBI_ACL 0x10 /* support ACL */
614 /* LL_SBI_RMT_CLIENT 0x40 remote client */
615 #define LL_SBI_MDS_CAPA 0x80 /* support mds capa, obsolete */
616 #define LL_SBI_OSS_CAPA 0x100 /* support oss capa, obsolete */
617 #define LL_SBI_LOCALFLOCK 0x200 /* Local flocks support by kernel */
618 #define LL_SBI_LRU_RESIZE 0x400 /* lru resize support */
619 #define LL_SBI_LAZYSTATFS 0x800 /* lazystatfs mount option */
620 /* LL_SBI_SOM_PREVIEW 0x1000 SOM preview mount option, obsolete */
621 #define LL_SBI_32BIT_API 0x2000 /* generate 32 bit inodes. */
622 #define LL_SBI_64BIT_HASH 0x4000 /* support 64-bits dir hash/offset */
623 #define LL_SBI_AGL_ENABLED 0x8000 /* enable agl */
624 #define LL_SBI_VERBOSE 0x10000 /* verbose mount/umount */
625 #define LL_SBI_LAYOUT_LOCK 0x20000 /* layout lock support */
626 #define LL_SBI_USER_FID2PATH 0x40000 /* allow fid2path by unprivileged users */
627 #define LL_SBI_XATTR_CACHE 0x80000 /* support for xattr cache */
628 #define LL_SBI_NOROOTSQUASH 0x100000 /* do not apply root squash */
629 #define LL_SBI_ALWAYS_PING 0x200000 /* always ping even if server
631 #define LL_SBI_FAST_READ 0x400000 /* fast read support */
632 #define LL_SBI_FILE_SECCTX 0x800000 /* set file security context at create */
633 /* LL_SBI_PIO 0x1000000 parallel IO support, introduced in
635 #define LL_SBI_TINY_WRITE 0x2000000 /* tiny write support */
636 #define LL_SBI_FILE_HEAT 0x4000000 /* file heat support */
637 #define LL_SBI_TEST_DUMMY_ENCRYPTION 0x8000000 /* test dummy encryption */
638 #define LL_SBI_ENCRYPT 0x10000000 /* client side encryption */
639 #define LL_SBI_FOREIGN_SYMLINK 0x20000000 /* foreign fake-symlink support */
640 /* foreign fake-symlink upcall registered */
641 #define LL_SBI_FOREIGN_SYMLINK_UPCALL 0x40000000
642 #define LL_SBI_PARALLEL_DIO 0x80000000 /* parallel (async) submission of
644 #define LL_SBI_FLAGS { \
672 "test_dummy_encryption", \
675 "foreign_symlink_upcall", \
679 /* This is embedded into llite super-blocks to keep track of connect
680 * flags (capabilities) supported by all imports given mount is
682 struct lustre_client_ocd {
683 /* This is conjunction of connect_flags across all imports
684 * (LOVs) this mount is connected to. This field is updated by
685 * cl_ocd_update() under ->lco_lock. */
687 struct mutex lco_lock;
688 struct obd_export *lco_md_exp;
689 struct obd_export *lco_dt_exp;
693 /* this protects pglist and ra_info. It isn't safe to
694 * grab from interrupt contexts */
696 spinlock_t ll_pp_extent_lock; /* pp_extent entry*/
697 spinlock_t ll_process_lock; /* ll_rw_process_info */
698 struct obd_uuid ll_sb_uuid;
699 struct obd_export *ll_md_exp;
700 struct obd_export *ll_dt_exp;
701 struct obd_device *ll_md_obd;
702 struct obd_device *ll_dt_obd;
703 struct dentry *ll_debugfs_entry;
704 struct lu_fid ll_root_fid; /* root object fid */
707 unsigned int ll_xattr_cache_enabled:1,
708 ll_xattr_cache_set:1, /* already set to 0/1 */
709 ll_client_common_fill_super_succeeded:1,
712 struct lustre_client_ocd ll_lco;
714 struct lprocfs_stats *ll_stats; /* lprocfs stats counter */
716 /* Used to track "unstable" pages on a client, and maintain a
717 * LRU list of clean pages. An "unstable" page is defined as
718 * any page which is sent to a server as part of a bulk request,
719 * but is uncommitted to stable storage. */
720 struct cl_client_cache *ll_cache;
722 struct lprocfs_stats *ll_ra_stats;
724 struct ll_ra_info ll_ra_info;
725 unsigned int ll_namelen;
726 const struct file_operations *ll_fop;
728 struct lu_site *ll_site;
729 struct cl_device *ll_cl;
731 struct ll_rw_extents_info ll_rw_extents_info;
732 int ll_extent_process_count;
733 struct ll_rw_process_info ll_rw_process_info[LL_PROCESS_HIST_MAX];
734 unsigned int ll_offset_process_count;
735 struct ll_rw_process_info ll_rw_offset_info[LL_OFFSET_HIST_MAX];
736 unsigned int ll_rw_offset_entry_count;
737 int ll_stats_track_id;
738 enum stats_track_type ll_stats_track_type;
741 /* metadata stat-ahead */
742 unsigned int ll_sa_running_max;/* max concurrent
743 * statahead instances */
744 unsigned int ll_sa_max; /* max statahead RPCs */
745 atomic_t ll_sa_total; /* statahead thread started
747 atomic_t ll_sa_wrong; /* statahead thread stopped for
749 atomic_t ll_sa_running; /* running statahead thread
751 atomic_t ll_agl_total; /* AGL thread started count */
753 dev_t ll_sdev_orig; /* save s_dev before assign for
756 struct root_squash_info ll_squash;
759 /* st_blksize returned by stat(2), when non-zero */
760 unsigned int ll_stat_blksize;
762 /* maximum relative age of cached statfs results */
763 unsigned int ll_statfs_max_age;
765 struct kset ll_kset; /* sysfs object */
766 struct completion ll_kobj_unregister;
769 unsigned int ll_heat_decay_weight;
770 unsigned int ll_heat_period_second;
772 /* Opens of the same inode before we start requesting open lock */
773 u32 ll_oc_thrsh_count;
775 /* Time in ms between last inode close and next open to be considered
776 * instant back to back and would trigger an open lock request
780 /* Time in ms after last file close that we no longer count prior opens*/
783 /* filesystem fsname */
784 char ll_fsname[LUSTRE_MAXFSNAME + 1];
786 /* Persistent Client Cache */
787 struct pcc_super ll_pcc_super;
789 /* to protect vs updates in all following foreign symlink fields */
790 struct rw_semaphore ll_foreign_symlink_sem;
791 /* foreign symlink path prefix */
792 char *ll_foreign_symlink_prefix;
793 /* full prefix size including leading '\0' */
794 size_t ll_foreign_symlink_prefix_size;
795 /* foreign symlink path upcall */
796 char *ll_foreign_symlink_upcall;
797 /* foreign symlink path upcall infos */
798 struct ll_foreign_symlink_upcall_item *ll_foreign_symlink_upcall_items;
799 /* foreign symlink path upcall nb infos */
800 unsigned int ll_foreign_symlink_upcall_nb_items;
803 #define SBI_DEFAULT_HEAT_DECAY_WEIGHT ((80 * 256 + 50) / 100)
804 #define SBI_DEFAULT_HEAT_PERIOD_SECOND (60)
806 #define SBI_DEFAULT_OPENCACHE_THRESHOLD_COUNT (5)
807 #define SBI_DEFAULT_OPENCACHE_THRESHOLD_MS (100) /* 0.1 second */
808 #define SBI_DEFAULT_OPENCACHE_THRESHOLD_MAX_MS (60000) /* 1 minute */
811 * per file-descriptor read-ahead data.
813 struct ll_readahead_state {
815 /* End byte that read(2) try to read. */
816 loff_t ras_last_read_end_bytes;
818 * number of bytes read after last read-ahead window reset. As window
819 * is reset on each seek, this is effectively a number of consecutive
820 * accesses. Maybe ->ras_accessed_in_window is better name.
822 * XXX nikita: window is also reset (by ras_update()) when Lustre
823 * believes that memory pressure evicts read-ahead pages. In that
824 * case, it probably doesn't make sense to expand window to
825 * PTLRPC_MAX_BRW_PAGES on the third access.
827 loff_t ras_consecutive_bytes;
829 * number of read requests after the last read-ahead window reset
830 * As window is reset on each seek, this is effectively the number
831 * on consecutive read request and is used to trigger read-ahead.
833 unsigned long ras_consecutive_requests;
835 * Parameters of current read-ahead window. Handled by
836 * ras_update(). On the initial access to the file or after a seek,
837 * window is reset to 0. After 3 consecutive accesses, window is
838 * expanded to PTLRPC_MAX_BRW_PAGES. Afterwards, window is enlarged by
839 * PTLRPC_MAX_BRW_PAGES chunks up to ->ra_max_pages.
841 pgoff_t ras_window_start_idx;
842 pgoff_t ras_window_pages;
844 /* Page index where min range read starts */
845 pgoff_t ras_range_min_start_idx;
846 /* Page index where mmap range read ends */
847 pgoff_t ras_range_max_end_idx;
848 /* number of mmap pages where last time detected */
849 pgoff_t ras_last_range_pages;
850 /* number of mmap range requests */
851 pgoff_t ras_range_requests;
854 * Optimal RPC size in pages.
855 * It decides how many pages will be sent for each read-ahead.
857 unsigned long ras_rpc_pages;
859 * Where next read-ahead should start at. This lies within read-ahead
860 * window. Read-ahead window is read in pieces rather than at once
861 * because: 1. lustre limits total number of pages under read-ahead by
862 * ->ra_max_pages (see ll_ra_count_get()), 2. client cannot read pages
863 * not covered by DLM lock.
865 pgoff_t ras_next_readahead_idx;
867 * Total number of ll_file_read requests issued, reads originating
868 * due to mmap are not counted in this total. This value is used to
869 * trigger full file read-ahead after multiple reads to a small file.
871 unsigned long ras_requests;
873 * The following 3 items are used for detecting the stride I/O
875 * In stride I/O mode,
876 * ...............|-----data-----|****gap*****|--------|******|....
877 * offset |-stride_bytes-|-stride_gap-|
878 * ras_stride_offset = offset;
879 * ras_stride_length = stride_bytes + stride_gap;
880 * ras_stride_bytes = stride_bytes;
881 * Note: all these three items are counted by bytes.
883 loff_t ras_stride_offset;
884 loff_t ras_stride_length;
885 loff_t ras_stride_bytes;
887 * number of consecutive stride request count, and it is similar as
888 * ras_consecutive_requests, but used for stride I/O mode.
889 * Note: only more than 2 consecutive stride request are detected,
890 * stride read-ahead will be enable
892 unsigned long ras_consecutive_stride_requests;
893 /* index of the last page that async readahead starts */
894 pgoff_t ras_async_last_readpage_idx;
895 /* whether we should increase readahead window */
896 bool ras_need_increase_window;
897 /* whether ra miss check should be skipped */
898 bool ras_no_miss_check;
901 struct ll_readahead_work {
902 /** File to readahead */
903 struct file *lrw_file;
904 pgoff_t lrw_start_idx;
907 /* async worker to handler read */
908 struct work_struct lrw_readahead_work;
909 char lrw_jobid[LUSTRE_JOBID_SIZE];
912 extern struct kmem_cache *ll_file_data_slab;
913 struct lustre_handle;
914 struct ll_file_data {
915 struct ll_readahead_state fd_ras;
916 struct ll_grouplock fd_grouplock;
920 /* openhandle if lease exists for this file.
921 * Borrow lli->lli_och_mutex to protect assignment */
922 struct obd_client_handle *fd_lease_och;
923 struct obd_client_handle *fd_och;
924 struct file *fd_file;
925 /* Indicate whether need to report failure when close.
926 * true: failure is known, not report again.
927 * false: unknown failure, should report. */
928 bool fd_write_failed;
929 bool ll_lock_no_expand;
930 rwlock_t fd_lock; /* protect lcc list */
931 struct list_head fd_lccs; /* list of ll_cl_context */
932 /* Used by mirrored file to lead IOs to a specific mirror, usually
933 * for mirror resync. 0 means default. */
934 __u32 fd_designated_mirror;
935 /* The layout version when resync starts. Resync I/O should carry this
936 * layout version for verification to OST objects */
937 __u32 fd_layout_version;
938 struct pcc_file fd_pcc_file;
941 void llite_tunables_unregister(void);
942 int llite_tunables_register(void);
944 static inline struct inode *ll_info2i(struct ll_inode_info *lli)
946 return &lli->lli_vfs_inode;
949 __u32 ll_i2suppgid(struct inode *i);
950 void ll_i2gids(__u32 *suppgids, struct inode *i1,struct inode *i2);
952 static inline int ll_need_32bit_api(struct ll_sb_info *sbi)
954 #if BITS_PER_LONG == 32
956 #elif defined(CONFIG_COMPAT)
957 if (unlikely(sbi->ll_flags & LL_SBI_32BIT_API))
960 # ifdef CONFIG_X86_X32
961 /* in_compat_syscall() returns true when called from a kthread
962 * and CONFIG_X86_X32 is enabled, which is wrong. So check
963 * whether the caller comes from a syscall (ie. not a kthread)
964 * before calling in_compat_syscall(). */
965 if (current->flags & PF_KTHREAD)
969 return unlikely(in_compat_syscall());
971 return unlikely(sbi->ll_flags & LL_SBI_32BIT_API);
975 static inline bool ll_sbi_has_fast_read(struct ll_sb_info *sbi)
977 return !!(sbi->ll_flags & LL_SBI_FAST_READ);
980 static inline bool ll_sbi_has_tiny_write(struct ll_sb_info *sbi)
982 return !!(sbi->ll_flags & LL_SBI_TINY_WRITE);
985 static inline bool ll_sbi_has_file_heat(struct ll_sb_info *sbi)
987 return !!(sbi->ll_flags & LL_SBI_FILE_HEAT);
990 static inline bool ll_sbi_has_foreign_symlink(struct ll_sb_info *sbi)
992 return !!(sbi->ll_flags & LL_SBI_FOREIGN_SYMLINK);
995 static inline bool ll_sbi_has_parallel_dio(struct ll_sb_info *sbi)
997 return !!(sbi->ll_flags & LL_SBI_PARALLEL_DIO);
1000 void ll_ras_enter(struct file *f, loff_t pos, size_t count);
1002 /* llite/lcommon_misc.c */
1003 int cl_ocd_update(struct obd_device *host, struct obd_device *watched,
1004 enum obd_notify_event ev, void *owner);
1005 int cl_get_grouplock(struct cl_object *obj, unsigned long gid, int nonblock,
1006 struct ll_grouplock *lg);
1007 void cl_put_grouplock(struct ll_grouplock *lg);
1009 /* llite/lproc_llite.c */
1010 int ll_debugfs_register_super(struct super_block *sb, const char *name);
1011 void ll_debugfs_unregister_super(struct super_block *sb);
1012 void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, long count);
1015 LPROC_LL_READ_BYTES,
1016 LPROC_LL_WRITE_BYTES,
1043 LPROC_LL_GETXATTR_HITS,
1045 LPROC_LL_REMOVEXATTR,
1046 LPROC_LL_INODE_PERM,
1048 LPROC_LL_INODE_OCOUNT,
1049 LPROC_LL_INODE_OPCLTM,
1050 LPROC_LL_FILE_OPCODES
1054 enum get_default_layout_type {
1055 GET_DEFAULT_LAYOUT_ROOT = 1,
1058 extern const struct file_operations ll_dir_operations;
1059 extern const struct inode_operations ll_dir_inode_operations;
1060 #ifdef HAVE_DIR_CONTEXT
1061 int ll_dir_read(struct inode *inode, __u64 *pos, struct md_op_data *op_data,
1062 struct dir_context *ctx);
1064 int ll_dir_read(struct inode *inode, __u64 *pos, struct md_op_data *op_data,
1065 void *cookie, filldir_t filldir);
1067 int ll_get_mdt_idx(struct inode *inode);
1068 int ll_get_mdt_idx_by_fid(struct ll_sb_info *sbi, const struct lu_fid *fid);
1069 struct page *ll_get_dir_page(struct inode *dir, struct md_op_data *op_data,
1071 void ll_release_page(struct inode *inode, struct page *page, bool remove);
1072 int quotactl_ioctl(struct ll_sb_info *sbi, struct if_quotactl *qctl);
1075 extern const struct inode_operations ll_special_inode_operations;
1077 struct inode *ll_iget(struct super_block *sb, ino_t hash,
1078 struct lustre_md *lic);
1079 int ll_test_inode_by_fid(struct inode *inode, void *opaque);
1080 int ll_md_blocking_ast(struct ldlm_lock *, struct ldlm_lock_desc *,
1081 void *data, int flag);
1082 struct dentry *ll_splice_alias(struct inode *inode, struct dentry *de);
1083 int ll_rmdir_entry(struct inode *dir, char *name, int namelen);
1084 void ll_update_times(struct ptlrpc_request *request, struct inode *inode);
1087 int ll_writepage(struct page *page, struct writeback_control *wbc);
1088 int ll_writepages(struct address_space *, struct writeback_control *wbc);
1089 int ll_readpage(struct file *file, struct page *page);
1090 int ll_io_read_page(const struct lu_env *env, struct cl_io *io,
1091 struct cl_page *page, struct file *file);
1092 void ll_readahead_init(struct inode *inode, struct ll_readahead_state *ras);
1093 int vvp_io_write_commit(const struct lu_env *env, struct cl_io *io);
1096 void ll_cl_add(struct file *file, const struct lu_env *env, struct cl_io *io,
1097 enum lcc_type type);
1098 void ll_cl_remove(struct file *file, const struct lu_env *env);
1099 struct ll_cl_context *ll_cl_find(struct file *file);
1101 extern const struct address_space_operations ll_aops;
1104 extern const struct inode_operations ll_file_inode_operations;
1105 const struct file_operations *ll_select_file_operations(struct ll_sb_info *sbi);
1106 extern int ll_have_md_lock(struct inode *inode, __u64 *bits,
1107 enum ldlm_mode l_req_mode);
1108 extern enum ldlm_mode ll_take_md_lock(struct inode *inode, __u64 bits,
1109 struct lustre_handle *lockh, __u64 flags,
1110 enum ldlm_mode mode);
1112 int ll_file_open(struct inode *inode, struct file *file);
1113 int ll_file_release(struct inode *inode, struct file *file);
1114 int ll_release_openhandle(struct dentry *, struct lookup_intent *);
1115 int ll_md_real_close(struct inode *inode, fmode_t fmode);
1116 void ll_track_file_opens(struct inode *inode);
1117 extern void ll_rw_stats_tally(struct ll_sb_info *sbi, pid_t pid,
1118 struct ll_file_data *file, loff_t pos,
1119 size_t count, int rw);
1120 #ifdef HAVE_INODEOPS_ENHANCED_GETATTR
1121 int ll_getattr(const struct path *path, struct kstat *stat,
1122 u32 request_mask, unsigned int flags);
1124 int ll_getattr(struct vfsmount *mnt, struct dentry *de, struct kstat *stat);
1126 int ll_getattr_dentry(struct dentry *de, struct kstat *stat, u32 request_mask,
1127 unsigned int flags, bool foreign);
1128 #ifdef CONFIG_LUSTRE_FS_POSIX_ACL
1129 struct posix_acl *ll_get_acl(struct inode *inode, int type);
1130 int ll_set_acl(struct inode *inode, struct posix_acl *acl, int type);
1131 #else /* !CONFIG_LUSTRE_FS_POSIX_ACL */
1132 #define ll_get_acl NULL
1133 #define ll_set_acl NULL
1134 #endif /* CONFIG_LUSTRE_FS_POSIX_ACL */
1136 static inline int ll_xflags_to_inode_flags(int xflags)
1138 return ((xflags & FS_XFLAG_SYNC) ? S_SYNC : 0) |
1139 ((xflags & FS_XFLAG_NOATIME) ? S_NOATIME : 0) |
1140 ((xflags & FS_XFLAG_APPEND) ? S_APPEND : 0) |
1141 ((xflags & FS_XFLAG_IMMUTABLE) ? S_IMMUTABLE : 0);
1144 static inline int ll_inode_flags_to_xflags(int inode_flags)
1146 return ((inode_flags & S_SYNC) ? FS_XFLAG_SYNC : 0) |
1147 ((inode_flags & S_NOATIME) ? FS_XFLAG_NOATIME : 0) |
1148 ((inode_flags & S_APPEND) ? FS_XFLAG_APPEND : 0) |
1149 ((inode_flags & S_IMMUTABLE) ? FS_XFLAG_IMMUTABLE : 0);
1152 int ll_migrate(struct inode *parent, struct file *file,
1153 struct lmv_user_md *lum, const char *name);
1154 int ll_get_fid_by_name(struct inode *parent, const char *name,
1155 int namelen, struct lu_fid *fid, struct inode **inode);
1156 int ll_inode_permission(struct inode *inode, int mask);
1157 int ll_ioctl_check_project(struct inode *inode, struct fsxattr *fa);
1158 int ll_ioctl_fsgetxattr(struct inode *inode, unsigned int cmd,
1160 int ll_ioctl_fssetxattr(struct inode *inode, unsigned int cmd,
1163 int ll_lov_setstripe_ea_info(struct inode *inode, struct dentry *dentry,
1164 __u64 flags, struct lov_user_md *lum,
1166 int ll_lov_getstripe_ea_info(struct inode *inode, const char *filename,
1167 struct lov_mds_md **lmm, int *lmm_size,
1168 struct ptlrpc_request **request);
1169 int ll_dir_setstripe(struct inode *inode, struct lov_user_md *lump,
1171 int ll_dir_getstripe_default(struct inode *inode, void **lmmp,
1172 int *lmm_size, struct ptlrpc_request **request,
1173 struct ptlrpc_request **root_request, u64 valid);
1174 int ll_dir_getstripe(struct inode *inode, void **plmm, int *plmm_size,
1175 struct ptlrpc_request **request, u64 valid);
1176 int ll_fsync(struct file *file, loff_t start, loff_t end, int data);
1177 int ll_merge_attr(const struct lu_env *env, struct inode *inode);
1178 int ll_fid2path(struct inode *inode, void __user *arg);
1179 int ll_data_version(struct inode *inode, __u64 *data_version, int flags);
1180 int ll_hsm_release(struct inode *inode);
1181 int ll_hsm_state_set(struct inode *inode, struct hsm_state_set *hss);
1182 void ll_io_set_mirror(struct cl_io *io, const struct file *file);
1184 /* llite/dcache.c */
1186 int ll_d_init(struct dentry *de);
1187 extern const struct dentry_operations ll_d_ops;
1188 void ll_intent_drop_lock(struct lookup_intent *);
1189 void ll_intent_release(struct lookup_intent *);
1190 void ll_prune_aliases(struct inode *inode);
1191 void ll_lookup_finish_locks(struct lookup_intent *it, struct dentry *dentry);
1192 int ll_revalidate_it_finish(struct ptlrpc_request *request,
1193 struct lookup_intent *it, struct dentry *de);
1195 /* llite/llite_lib.c */
1196 extern const struct super_operations lustre_super_operations;
1198 void ll_lli_init(struct ll_inode_info *lli);
1199 int ll_fill_super(struct super_block *sb);
1200 void ll_put_super(struct super_block *sb);
1201 void ll_kill_super(struct super_block *sb);
1202 struct inode *ll_inode_from_resource_lock(struct ldlm_lock *lock);
1203 void ll_dir_clear_lsm_md(struct inode *inode);
1204 void ll_clear_inode(struct inode *inode);
1205 int ll_setattr_raw(struct dentry *dentry, struct iattr *attr,
1206 enum op_xvalid xvalid, bool hsm_import);
1207 int ll_setattr(struct dentry *de, struct iattr *attr);
1208 int ll_statfs(struct dentry *de, struct kstatfs *sfs);
1209 int ll_statfs_internal(struct ll_sb_info *sbi, struct obd_statfs *osfs,
1211 int ll_update_inode(struct inode *inode, struct lustre_md *md);
1212 void ll_update_inode_flags(struct inode *inode, unsigned int ext_flags);
1213 int ll_read_inode2(struct inode *inode, void *opaque);
1214 void ll_truncate_inode_pages_final(struct inode *inode);
1215 void ll_delete_inode(struct inode *inode);
1216 int ll_iocontrol(struct inode *inode, struct file *file,
1217 unsigned int cmd, unsigned long arg);
1218 int ll_flush_ctx(struct inode *inode);
1219 void ll_umount_begin(struct super_block *sb);
1220 int ll_remount_fs(struct super_block *sb, int *flags, char *data);
1221 int ll_show_options(struct seq_file *seq, struct dentry *dentry);
1222 void ll_dirty_page_discard_warn(struct page *page, int ioret);
1223 int ll_prep_inode(struct inode **inode, struct req_capsule *pill,
1224 struct super_block *sb, struct lookup_intent *it);
1225 int ll_obd_statfs(struct inode *inode, void __user *arg);
1226 int ll_get_max_mdsize(struct ll_sb_info *sbi, int *max_mdsize);
1227 int ll_get_default_mdsize(struct ll_sb_info *sbi, int *default_mdsize);
1228 int ll_set_default_mdsize(struct ll_sb_info *sbi, int default_mdsize);
1230 void ll_unlock_md_op_lsm(struct md_op_data *op_data);
1231 struct md_op_data *ll_prep_md_op_data(struct md_op_data *op_data,
1232 struct inode *i1, struct inode *i2,
1233 const char *name, size_t namelen,
1234 __u32 mode, enum md_op_code opc,
1236 void ll_finish_md_op_data(struct md_op_data *op_data);
1237 int ll_get_obd_name(struct inode *inode, unsigned int cmd, unsigned long arg);
1238 void ll_compute_rootsquash_state(struct ll_sb_info *sbi);
1239 ssize_t ll_copy_user_md(const struct lov_user_md __user *md,
1240 struct lov_user_md **kbuf);
1241 void ll_open_cleanup(struct super_block *sb, struct req_capsule *pill);
1243 void ll_dom_finish_open(struct inode *inode, struct ptlrpc_request *req);
1245 /* Compute expected user md size when passing in a md from user space */
1246 static inline ssize_t ll_lov_user_md_size(const struct lov_user_md *lum)
1248 switch (lum->lmm_magic) {
1249 case LOV_USER_MAGIC_V1:
1250 return sizeof(struct lov_user_md_v1);
1251 case LOV_USER_MAGIC_V3:
1252 return sizeof(struct lov_user_md_v3);
1253 case LOV_USER_MAGIC_SPECIFIC:
1254 if (lum->lmm_stripe_count > LOV_MAX_STRIPE_COUNT)
1257 return lov_user_md_size(lum->lmm_stripe_count,
1258 LOV_USER_MAGIC_SPECIFIC);
1259 case LOV_USER_MAGIC_COMP_V1:
1260 return ((struct lov_comp_md_v1 *)lum)->lcm_size;
1261 case LOV_USER_MAGIC_FOREIGN:
1262 return foreign_size(lum);
1268 /* llite/llite_nfs.c */
1269 extern const struct export_operations lustre_export_operations;
1270 __u32 get_uuid2int(const char *name, int len);
1271 struct inode *search_inode_for_lustre(struct super_block *sb,
1272 const struct lu_fid *fid);
1273 int ll_dir_get_parent_fid(struct inode *dir, struct lu_fid *parent_fid);
1275 /* llite/symlink.c */
1276 extern const struct inode_operations ll_fast_symlink_inode_operations;
1279 * IO arguments for various VFS I/O interfaces.
1281 struct vvp_io_args {
1282 /** normal/sendfile/splice */
1285 struct kiocb *via_iocb;
1286 struct iov_iter *via_iter;
1296 struct ll_cl_context {
1297 struct list_head lcc_list;
1299 const struct lu_env *lcc_env;
1300 struct cl_io *lcc_io;
1301 struct cl_page *lcc_page;
1302 enum lcc_type lcc_type;
1305 struct ll_thread_info {
1306 struct vvp_io_args lti_args;
1307 struct ra_io_arg lti_ria;
1308 struct ll_cl_context lti_io_ctx;
1311 extern struct lu_context_key ll_thread_key;
1313 static inline struct ll_thread_info *ll_env_info(const struct lu_env *env)
1315 struct ll_thread_info *lti;
1317 lti = lu_context_key_get(&env->le_ctx, &ll_thread_key);
1318 LASSERT(lti != NULL);
1323 static inline struct vvp_io_args *ll_env_args(const struct lu_env *env)
1325 return &ll_env_info(env)->lti_args;
1328 void ll_io_init(struct cl_io *io, struct file *file, enum cl_io_type iot,
1329 struct vvp_io_args *args);
1331 /* llite/llite_mmap.c */
1333 int ll_file_mmap(struct file * file, struct vm_area_struct * vma);
1334 void policy_from_vma(union ldlm_policy_data *policy, struct vm_area_struct *vma,
1335 unsigned long addr, size_t count);
1336 struct vm_area_struct *our_vma(struct mm_struct *mm, unsigned long addr,
1339 #define ll_s2sbi(sb) (s2lsi(sb)->lsi_llsbi)
1341 /* don't need an addref as the sb_info should be holding one */
1342 static inline struct obd_export *ll_s2dtexp(struct super_block *sb)
1344 return ll_s2sbi(sb)->ll_dt_exp;
1347 /* don't need an addref as the sb_info should be holding one */
1348 static inline struct obd_export *ll_s2mdexp(struct super_block *sb)
1350 return ll_s2sbi(sb)->ll_md_exp;
1353 static inline struct client_obd *sbi2mdc(struct ll_sb_info *sbi)
1355 struct obd_device *obd = sbi->ll_md_exp->exp_obd;
1361 // FIXME: replace the name of this with LL_SB to conform to kernel stuff
1362 static inline struct ll_sb_info *ll_i2sbi(struct inode *inode)
1364 return ll_s2sbi(inode->i_sb);
1367 static inline struct obd_export *ll_i2dtexp(struct inode *inode)
1369 return ll_s2dtexp(inode->i_sb);
1372 static inline struct obd_export *ll_i2mdexp(struct inode *inode)
1374 return ll_s2mdexp(inode->i_sb);
1377 static inline struct lu_fid *ll_inode2fid(struct inode *inode)
1381 LASSERT(inode != NULL);
1382 fid = &ll_i2info(inode)->lli_fid;
1387 static inline bool ll_dir_striped(struct inode *inode)
1390 return S_ISDIR(inode->i_mode) &&
1391 lmv_dir_striped(ll_i2info(inode)->lli_lsm_md);
1394 static inline loff_t ll_file_maxbytes(struct inode *inode)
1396 struct cl_object *obj = ll_i2info(inode)->lli_clob;
1399 return MAX_LFS_FILESIZE;
1401 return min_t(loff_t, cl_object_maxbytes(obj), MAX_LFS_FILESIZE);
1405 extern const struct xattr_handler *ll_xattr_handlers[];
1407 #define XATTR_USER_T 1
1408 #define XATTR_TRUSTED_T 2
1409 #define XATTR_SECURITY_T 3
1410 #define XATTR_ACL_ACCESS_T 4
1411 #define XATTR_ACL_DEFAULT_T 5
1412 #define XATTR_LUSTRE_T 6
1413 #define XATTR_OTHER_T 7
1415 ssize_t ll_listxattr(struct dentry *dentry, char *buffer, size_t size);
1416 int ll_xattr_list(struct inode *inode, const char *name, int type,
1417 void *buffer, size_t size, u64 valid);
1418 const struct xattr_handler *get_xattr_type(const char *name);
1421 * Common IO arguments for various VFS I/O interfaces.
1423 int cl_sb_init(struct super_block *sb);
1424 int cl_sb_fini(struct super_block *sb);
1426 enum ras_update_flags {
1430 void ll_ra_count_put(struct ll_sb_info *sbi, unsigned long len);
1431 void ll_ra_stats_inc(struct inode *inode, enum ra_stat which);
1435 #define LL_SA_RPC_MIN 2
1436 #define LL_SA_RPC_DEF 32
1437 #define LL_SA_RPC_MAX 512
1439 /* XXX: If want to support more concurrent statahead instances,
1440 * please consider to decentralize the RPC lists attached
1441 * on related import, such as imp_{sending,delayed}_list.
1443 #define LL_SA_RUNNING_MAX 256
1444 #define LL_SA_RUNNING_DEF 16
1446 #define LL_SA_CACHE_BIT 5
1447 #define LL_SA_CACHE_SIZE (1 << LL_SA_CACHE_BIT)
1448 #define LL_SA_CACHE_MASK (LL_SA_CACHE_SIZE - 1)
1450 /* per inode struct, for dir only */
1451 struct ll_statahead_info {
1452 struct dentry *sai_dentry;
1453 atomic_t sai_refcount; /* when access this struct, hold
1455 unsigned int sai_max; /* max ahead of lookup */
1456 __u64 sai_sent; /* stat requests sent count */
1457 __u64 sai_replied; /* stat requests which received
1459 __u64 sai_index; /* index of statahead entry */
1460 __u64 sai_index_wait; /* index of entry which is the
1461 * caller is waiting for */
1462 __u64 sai_hit; /* hit count */
1463 __u64 sai_miss; /* miss count:
1464 * for "ls -al" case, includes
1465 * hidden dentry miss;
1466 * for "ls -l" case, it does not
1467 * include hidden dentry miss.
1468 * "sai_miss_hidden" is used for
1471 unsigned int sai_consecutive_miss; /* consecutive miss */
1472 unsigned int sai_miss_hidden;/* "ls -al", but first dentry
1473 * is not a hidden one */
1474 unsigned int sai_skip_hidden;/* skipped hidden dentry count
1476 unsigned int sai_ls_all:1; /* "ls -al", do stat-ahead for
1478 wait_queue_head_t sai_waitq; /* stat-ahead wait queue */
1479 struct task_struct *sai_task; /* stat-ahead thread */
1480 struct task_struct *sai_agl_task; /* AGL thread */
1481 struct list_head sai_entries; /* completed entries */
1482 struct list_head sai_agls; /* AGLs to be sent */
1483 struct list_head sai_cache[LL_SA_CACHE_SIZE];
1484 spinlock_t sai_cache_lock[LL_SA_CACHE_SIZE];
1485 atomic_t sai_cache_count; /* entry count in cache */
1488 int ll_revalidate_statahead(struct inode *dir, struct dentry **dentry,
1490 int ll_start_statahead(struct inode *dir, struct dentry *dentry, bool agl);
1491 void ll_authorize_statahead(struct inode *dir, void *key);
1492 void ll_deauthorize_statahead(struct inode *dir, void *key);
1495 blkcnt_t dirty_cnt(struct inode *inode);
1497 int cl_glimpse_size0(struct inode *inode, int agl);
1498 int cl_glimpse_lock(const struct lu_env *env, struct cl_io *io,
1499 struct inode *inode, struct cl_object *clob, int agl);
1501 static inline int cl_glimpse_size(struct inode *inode)
1503 return cl_glimpse_size0(inode, 0);
1506 /* AGL is 'asychronous glimpse lock', which is a speculative lock taken as
1507 * part of statahead */
1508 static inline int cl_agl(struct inode *inode)
1510 return cl_glimpse_size0(inode, 1);
1513 int ll_file_lock_ahead(struct file *file, struct llapi_lu_ladvise *ladvise);
1515 int cl_io_get(struct inode *inode, struct lu_env **envout,
1516 struct cl_io **ioout, __u16 *refcheck);
1518 static inline int ll_glimpse_size(struct inode *inode)
1520 struct ll_inode_info *lli = ll_i2info(inode);
1523 down_read(&lli->lli_glimpse_sem);
1524 rc = cl_glimpse_size(inode);
1525 lli->lli_glimpse_time = ktime_get();
1526 up_read(&lli->lli_glimpse_sem);
1530 /* dentry may statahead when statahead is enabled and current process has opened
1531 * parent directory, and this dentry hasn't accessed statahead cache before */
1533 dentry_may_statahead(struct inode *dir, struct dentry *dentry)
1535 struct ll_inode_info *lli;
1536 struct ll_dentry_data *ldd;
1538 if (ll_i2sbi(dir)->ll_sa_max == 0)
1541 lli = ll_i2info(dir);
1543 /* statahead is not allowed for this dir, there may be three causes:
1544 * 1. dir is not opened.
1545 * 2. statahead hit ratio is too low.
1546 * 3. previous stat started statahead thread failed. */
1547 if (!lli->lli_sa_enabled)
1550 /* not the same process, don't statahead */
1551 if (lli->lli_opendir_pid != current->pid)
1555 * When stating a dentry, kernel may trigger 'revalidate' or 'lookup'
1556 * multiple times, eg. for 'getattr', 'getxattr' and etc.
1557 * For patchless client, lookup intent is not accurate, which may
1558 * misguide statahead. For example:
1559 * The 'revalidate' call for 'getattr' and 'getxattr' of a dentry will
1560 * have the same intent -- IT_GETATTR, while one dentry should access
1561 * statahead cache once, otherwise statahead windows is messed up.
1562 * The solution is as following:
1563 * Assign 'lld_sa_generation' with 'lli_sa_generation' when a dentry
1564 * IT_GETATTR for the first time, and subsequent IT_GETATTR will
1565 * bypass interacting with statahead cache by checking
1566 * 'lld_sa_generation == lli->lli_sa_generation'.
1568 ldd = ll_d2d(dentry);
1569 if (ldd != NULL && lli->lli_sa_generation &&
1570 ldd->lld_sa_generation == lli->lli_sa_generation)
1576 int cl_sync_file_range(struct inode *inode, loff_t start, loff_t end,
1577 enum cl_fsync_mode mode, int ignore_layout);
1579 static inline int ll_file_nolock(const struct file *file)
1581 struct ll_file_data *fd = file->private_data;
1582 struct inode *inode = file_inode((struct file *)file);
1584 LASSERT(fd != NULL);
1585 return ((fd->fd_flags & LL_FILE_IGNORE_LOCK) ||
1586 (ll_i2sbi(inode)->ll_flags & LL_SBI_NOLCK));
1589 static inline void ll_set_lock_data(struct obd_export *exp, struct inode *inode,
1590 struct lookup_intent *it, __u64 *bits)
1592 if (!it->it_lock_set) {
1593 struct lustre_handle handle;
1595 /* If this inode is a remote object, it will get two
1596 * separate locks in different namespaces, Master MDT,
1597 * where the name entry is, will grant LOOKUP lock,
1598 * remote MDT, where the object is, will grant
1599 * UPDATE|PERM lock. The inode will be attched to both
1600 * LOOKUP and PERM locks, so revoking either locks will
1601 * case the dcache being cleared */
1602 if (it->it_remote_lock_mode) {
1603 handle.cookie = it->it_remote_lock_handle;
1604 CDEBUG(D_DLMTRACE, "setting l_data to inode "DFID
1605 "(%p) for remote lock %#llx\n",
1606 PFID(ll_inode2fid(inode)), inode,
1608 md_set_lock_data(exp, &handle, inode, NULL);
1611 handle.cookie = it->it_lock_handle;
1613 CDEBUG(D_DLMTRACE, "setting l_data to inode "DFID"(%p)"
1614 " for lock %#llx\n",
1615 PFID(ll_inode2fid(inode)), inode, handle.cookie);
1617 md_set_lock_data(exp, &handle, inode, &it->it_lock_bits);
1618 it->it_lock_set = 1;
1622 *bits = it->it_lock_bits;
1625 static inline int d_lustre_invalid(const struct dentry *dentry)
1627 struct ll_dentry_data *lld = ll_d2d(dentry);
1629 return (lld == NULL) || lld->lld_invalid;
1632 static inline void __d_lustre_invalidate(struct dentry *dentry)
1634 struct ll_dentry_data *lld = ll_d2d(dentry);
1637 lld->lld_invalid = 1;
1641 * Mark dentry INVALID, if dentry refcount is zero (this is normally case for
1642 * ll_md_blocking_ast), it will be pruned by ll_prune_aliases() and
1643 * ll_prune_negative_children(); otherwise dput() of the last refcount will
1644 * unhash this dentry and kill it.
1646 static inline void d_lustre_invalidate(struct dentry *dentry)
1648 CDEBUG(D_DENTRY, "invalidate dentry %pd (%p) parent %p inode %p refc %d\n",
1650 dentry->d_parent, dentry->d_inode, ll_d_count(dentry));
1652 spin_lock(&dentry->d_lock);
1653 __d_lustre_invalidate(dentry);
1654 spin_unlock(&dentry->d_lock);
1657 static inline void d_lustre_revalidate(struct dentry *dentry)
1659 spin_lock(&dentry->d_lock);
1660 LASSERT(ll_d2d(dentry) != NULL);
1661 ll_d2d(dentry)->lld_invalid = 0;
1662 spin_unlock(&dentry->d_lock);
1665 static inline dev_t ll_compat_encode_dev(dev_t dev)
1667 /* The compat_sys_*stat*() syscalls will fail unless the
1668 * device majors and minors are both less than 256. Note that
1669 * the value returned here will be passed through
1670 * old_encode_dev() in cp_compat_stat(). And so we are not
1671 * trying to return a valid compat (u16) device number, just
1672 * one that will pass the old_valid_dev() check. */
1674 return MKDEV(MAJOR(dev) & 0xff, MINOR(dev) & 0xff);
1677 int ll_layout_conf(struct inode *inode, const struct cl_object_conf *conf);
1678 int ll_layout_refresh(struct inode *inode, __u32 *gen);
1679 int ll_layout_restore(struct inode *inode, loff_t start, __u64 length);
1680 int ll_layout_write_intent(struct inode *inode, enum layout_intent_opc opc,
1681 struct lu_extent *ext);
1683 int ll_xattr_init(void);
1684 void ll_xattr_fini(void);
1686 int ll_page_sync_io(const struct lu_env *env, struct cl_io *io,
1687 struct cl_page *page, enum cl_req_type crt);
1689 int ll_getparent(struct file *file, struct getparent __user *arg);
1692 int cl_setattr_ost(struct cl_object *obj, const struct iattr *attr,
1693 enum op_xvalid xvalid, unsigned int attr_flags);
1695 extern struct lu_env *cl_inode_fini_env;
1696 extern __u16 cl_inode_fini_refcheck;
1698 int cl_file_inode_init(struct inode *inode, struct lustre_md *md);
1699 void cl_inode_fini(struct inode *inode);
1701 u64 cl_fid_build_ino(const struct lu_fid *fid, int api32);
1702 u32 cl_fid_build_gen(const struct lu_fid *fid);
1704 static inline struct pcc_super *ll_i2pccs(struct inode *inode)
1706 return &ll_i2sbi(inode)->ll_pcc_super;
1709 static inline struct pcc_super *ll_info2pccs(struct ll_inode_info *lli)
1711 return ll_i2pccs(ll_info2i(lli));
1714 #ifdef HAVE_LUSTRE_CRYPTO
1716 extern const struct llcrypt_operations lustre_cryptops;
1718 /* llite/llite_foreign.c */
1719 int ll_manage_foreign(struct inode *inode, struct lustre_md *lmd);
1720 bool ll_foreign_is_openable(struct dentry *dentry, unsigned int flags);
1721 bool ll_foreign_is_removable(struct dentry *dentry, bool unset);
1723 #endif /* LLITE_INTERNAL_H */