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 <lustre_compat.h>
47 #include <lustre_crypto.h>
48 #include <range_lock.h>
50 #include "vvp_internal.h"
52 #include "foreign_symlink.h"
58 #ifndef HAVE_VM_FAULT_RETRY
59 #define VM_FAULT_RETRY 0
62 /* Kernel 3.1 kills LOOKUP_CONTINUE, LOOKUP_PARENT is equivalent to it.
63 * seem kernel commit 49084c3bb2055c401f3493c13edae14d49128ca0 */
64 #ifndef LOOKUP_CONTINUE
65 #define LOOKUP_CONTINUE LOOKUP_PARENT
68 /** Only used on client-side for indicating the tail of dir hash/offset. */
69 #define LL_DIR_END_OFF 0x7fffffffffffffffULL
70 #define LL_DIR_END_OFF_32BIT 0x7fffffffUL
72 /* 4UL * 1024 * 1024 */
73 #define LL_MAX_BLKSIZE_BITS 22
75 #define LL_IT2STR(it) ((it) ? ldlm_it2str((it)->it_op) : "0")
77 #define TIMES_SET_FLAGS (ATTR_MTIME_SET | ATTR_ATIME_SET | ATTR_TIMES_SET)
79 struct ll_dentry_data {
80 unsigned int lld_sa_generation;
81 unsigned int lld_invalid:1;
82 unsigned int lld_nfs_dentry:1;
83 struct rcu_head lld_rcu_head;
86 #define ll_d2d(de) ((struct ll_dentry_data*)((de)->d_fsdata))
88 #define LLI_INODE_MAGIC 0x111d0de5
89 #define LLI_INODE_DEAD 0xdeadd00d
91 struct ll_getname_data {
92 #ifdef HAVE_DIR_CONTEXT
93 struct dir_context ctx;
95 char *lgd_name; /* points to a buffer with NAME_MAX+1 size */
96 struct lu_fid lgd_fid; /* target fid we are looking for */
97 int lgd_found; /* inode matched? */
100 struct ll_grouplock {
101 struct lu_env *lg_env;
103 struct cl_lock *lg_lock;
104 unsigned long lg_gid;
107 /* See comment on trunc_sem_down_read_nowait */
108 struct ll_trunc_sem {
109 /* when positive, this is a count of readers, when -1, it indicates
110 * the semaphore is held for write, and 0 is unlocked
112 atomic_t ll_trunc_readers;
113 /* this tracks a count of waiting writers */
114 atomic_t ll_trunc_waiters;
117 struct ll_inode_info {
118 __u32 lli_inode_magic;
121 volatile unsigned long lli_flags;
122 struct posix_acl *lli_posix_acl;
124 /* identifying fields for both metadata and data stacks. */
125 struct lu_fid lli_fid;
126 /* master inode fid for stripe directory */
127 struct lu_fid lli_pfid;
129 /* We need all three because every inode may be opened in different
131 struct obd_client_handle *lli_mds_read_och;
132 struct obd_client_handle *lli_mds_write_och;
133 struct obd_client_handle *lli_mds_exec_och;
134 __u64 lli_open_fd_read_count;
135 __u64 lli_open_fd_write_count;
136 __u64 lli_open_fd_exec_count;
138 /* Number of times this inode was opened */
139 u64 lli_open_fd_count;
140 /* When last close was performed on this inode */
141 ktime_t lli_close_fd_time;
143 /* Protects access to och pointers and their usage counters */
144 struct mutex lli_och_mutex;
146 struct inode lli_vfs_inode;
148 /* the most recent timestamps obtained from mds */
153 spinlock_t lli_agl_lock;
155 /* Try to make the d::member and f::member are aligned. Before using
156 * these members, make clear whether it is directory or not. */
160 /* metadata statahead */
161 /* since parent-child threads can share the same @file
162 * struct, "opendir_key" is the token when dir close for
163 * case of parent exit before child -- it is me should
164 * cleanup the dir readahead. */
165 void *lli_opendir_key;
166 struct ll_statahead_info *lli_sai;
167 /* protect statahead stuff. */
168 spinlock_t lli_sa_lock;
169 /* "opendir_pid" is the token when lookup/revalid
170 * -- I am the owner of dir statahead. */
171 pid_t lli_opendir_pid;
172 /* directory depth to ROOT */
173 unsigned short lli_depth;
174 /* stat will try to access statahead entries or start
175 * statahead if this flag is set, and this flag will be
176 * set upon dir open, and cleared when dir is closed,
177 * statahead hit ratio is too low, or start statahead
179 unsigned short lli_sa_enabled:1;
180 /* generation for statahead */
181 unsigned int lli_sa_generation;
182 /* rw lock protects lli_lsm_md */
183 struct rw_semaphore lli_lsm_sem;
184 /* directory stripe information */
185 struct lmv_stripe_md *lli_lsm_md;
186 /* directory default LMV */
187 struct lmv_stripe_md *lli_default_lsm_md;
190 /* for non-directory */
192 struct mutex lli_size_mutex;
193 char *lli_symlink_name;
194 struct ll_trunc_sem lli_trunc_sem;
195 struct range_lock_tree lli_write_tree;
196 struct mutex lli_setattr_mutex;
198 struct rw_semaphore lli_glimpse_sem;
199 ktime_t lli_glimpse_time;
200 struct list_head lli_agl_list;
203 /* for writepage() only to communicate to fsync */
206 /* protect the file heat fields */
207 spinlock_t lli_heat_lock;
208 __u32 lli_heat_flags;
209 struct obd_heat_instance lli_heat_instances[OBD_HEAT_COUNT];
212 * Whenever a process try to read/write the file, the
213 * jobid of the process will be saved here, and it'll
214 * be packed into the write PRC when flush later.
216 * So the read/write statistics for jobid will not be
217 * accurate if the file is shared by different jobs.
219 char lli_jobid[LUSTRE_JOBID_SIZE];
221 struct mutex lli_pcc_lock;
222 enum lu_pcc_state_flags lli_pcc_state;
224 * @lli_pcc_generation saves the gobal PCC generation
225 * when the file was successfully attached into PCC.
226 * The flags of the PCC dataset are saved in
228 * The gobal PCC generation will be increased when add
229 * or delete a PCC backend, or change the configuration
230 * parameters for PCC.
231 * If @lli_pcc_generation is same as the gobal PCC
232 * generation, we can use the saved flags of the PCC
233 * dataset to determine whether need to try auto attach
236 __u64 lli_pcc_generation;
237 enum pcc_dataset_flags lli_pcc_dsflags;
238 struct pcc_inode *lli_pcc_inode;
240 struct mutex lli_group_mutex;
241 __u64 lli_group_users;
242 unsigned long lli_group_gid;
244 __u64 lli_attr_valid;
246 __u64 lli_lazyblocks;
250 /* XXX: For following frequent used members, although they maybe special
251 * used for non-directory object, it is some time-wasting to check
252 * whether the object is directory or not before using them. On the
253 * other hand, currently, sizeof(f) > sizeof(d), it cannot reduce
254 * the "ll_inode_info" size even if moving those members into u.f.
255 * So keep them out side.
257 * In the future, if more members are added only for directory,
258 * some of the following members can be moved into u.f.
260 struct cl_object *lli_clob;
262 /* mutex to request for layout lock exclusively. */
263 struct mutex lli_layout_mutex;
264 /* Layout version, protected by lli_layout_lock */
265 __u32 lli_layout_gen;
266 spinlock_t lli_layout_lock;
268 __u32 lli_projid; /* project id */
270 struct rw_semaphore lli_xattrs_list_rwsem;
271 struct mutex lli_xattrs_enq_lock;
272 struct list_head lli_xattrs; /* ll_xattr_entry->xe_list */
273 struct list_head lli_lccs; /* list of ll_cl_context */
276 static inline void ll_trunc_sem_init(struct ll_trunc_sem *sem)
278 atomic_set(&sem->ll_trunc_readers, 0);
279 atomic_set(&sem->ll_trunc_waiters, 0);
282 /* This version of down read ignores waiting writers, meaning if the semaphore
283 * is already held for read, this down_read will 'join' that reader and also
284 * take the semaphore.
286 * This lets us avoid an unusual deadlock.
288 * We must take lli_trunc_sem in read mode on entry in to various i/o paths
289 * in Lustre, in order to exclude truncates. Some of these paths then need to
290 * take the mmap_lock, while still holding the trunc_sem. The problem is that
291 * page faults hold the mmap_lock when calling in to Lustre, and then must also
292 * take the trunc_sem to exclude truncate.
294 * This means the locking order for trunc_sem and mmap_lock is sometimes AB,
295 * sometimes BA. This is almost OK because in both cases, we take the trunc
296 * sem for read, so it doesn't block.
298 * However, if a write mode user (truncate, a setattr op) arrives in the
299 * middle of this, the second reader on the truncate_sem will wait behind that
302 * So we have, on our truncate sem, in order (where 'reader' and 'writer' refer
303 * to the mode in which they take the semaphore):
304 * reader (holding mmap_lock, needs truncate_sem)
306 * reader (holding truncate sem, waiting for mmap_lock)
308 * And so the readers deadlock.
310 * The solution is this modified semaphore, where this down_read ignores
311 * waiting write operations, and all waiters are woken up at once, so readers
312 * using down_read_nowait cannot get stuck behind waiting writers, regardless
313 * of the order they arrived in.
315 * down_read_nowait is only used in the page fault case, where we already hold
316 * the mmap_lock. This is because otherwise repeated read and write operations
317 * (which take the truncate sem) could prevent a truncate from ever starting.
318 * This could still happen with page faults, but without an even more complex
319 * mechanism, this is unavoidable.
323 static inline void trunc_sem_down_read_nowait(struct ll_trunc_sem *sem)
325 wait_var_event(&sem->ll_trunc_readers,
326 atomic_inc_unless_negative(&sem->ll_trunc_readers));
329 static inline void trunc_sem_down_read(struct ll_trunc_sem *sem)
331 wait_var_event(&sem->ll_trunc_readers,
332 atomic_read(&sem->ll_trunc_waiters) == 0 &&
333 atomic_inc_unless_negative(&sem->ll_trunc_readers));
336 static inline void trunc_sem_up_read(struct ll_trunc_sem *sem)
338 if (atomic_dec_return(&sem->ll_trunc_readers) == 0 &&
339 atomic_read(&sem->ll_trunc_waiters))
340 wake_up_var(&sem->ll_trunc_readers);
343 static inline void trunc_sem_down_write(struct ll_trunc_sem *sem)
345 atomic_inc(&sem->ll_trunc_waiters);
346 wait_var_event(&sem->ll_trunc_readers,
347 atomic_cmpxchg(&sem->ll_trunc_readers, 0, -1) == 0);
348 atomic_dec(&sem->ll_trunc_waiters);
351 static inline void trunc_sem_up_write(struct ll_trunc_sem *sem)
353 atomic_set(&sem->ll_trunc_readers, 0);
354 /* match the smp_mb() in wait_var_event()->prepare_to_wait() */
356 wake_up_var(&sem->ll_trunc_readers);
359 #ifdef CONFIG_LUSTRE_FS_POSIX_ACL
360 static inline void lli_clear_acl(struct ll_inode_info *lli)
362 if (lli->lli_posix_acl) {
363 posix_acl_release(lli->lli_posix_acl);
364 lli->lli_posix_acl = NULL;
368 static inline void lli_replace_acl(struct ll_inode_info *lli,
369 struct lustre_md *md)
371 write_lock(&lli->lli_lock);
372 if (lli->lli_posix_acl)
373 posix_acl_release(lli->lli_posix_acl);
374 lli->lli_posix_acl = md->posix_acl;
375 write_unlock(&lli->lli_lock);
378 static inline void lli_clear_acl(struct ll_inode_info *lli)
382 static inline void lli_replace_acl(struct ll_inode_info *lli,
383 struct lustre_md *md)
388 static inline __u32 ll_layout_version_get(struct ll_inode_info *lli)
392 spin_lock(&lli->lli_layout_lock);
393 gen = lli->lli_layout_gen;
394 spin_unlock(&lli->lli_layout_lock);
399 static inline void ll_layout_version_set(struct ll_inode_info *lli, __u32 gen)
401 spin_lock(&lli->lli_layout_lock);
402 lli->lli_layout_gen = gen;
403 spin_unlock(&lli->lli_layout_lock);
407 /* File data is modified. */
408 LLIF_DATA_MODIFIED = 0,
409 /* File is being restored */
410 LLIF_FILE_RESTORING = 1,
411 /* Xattr cache is attached to the file */
412 LLIF_XATTR_CACHE = 2,
413 /* Project inherit */
414 LLIF_PROJECT_INHERIT = 3,
415 /* update atime from MDS even if it's older than local inode atime. */
416 LLIF_UPDATE_ATIME = 4,
417 /* foreign file/dir can be unlinked unconditionnaly */
418 LLIF_FOREIGN_REMOVABLE = 5,
419 /* setting encryption context in progress */
420 LLIF_SET_ENC_CTX = 6,
421 /* Xattr cache is filled */
422 LLIF_XATTR_CACHE_FILLED = 7,
426 int ll_xattr_cache_destroy(struct inode *inode);
427 int ll_xattr_cache_empty(struct inode *inode);
429 int ll_xattr_cache_get(struct inode *inode,
435 int ll_xattr_cache_insert(struct inode *inode,
440 static inline bool obd_connect_has_secctx(struct obd_connect_data *data)
442 #ifdef CONFIG_SECURITY
443 return data->ocd_connect_flags & OBD_CONNECT_FLAGS2 &&
444 data->ocd_connect_flags2 & OBD_CONNECT2_FILE_SECCTX;
450 static inline void obd_connect_set_secctx(struct obd_connect_data *data)
452 #ifdef CONFIG_SECURITY
453 data->ocd_connect_flags2 |= OBD_CONNECT2_FILE_SECCTX;
457 int ll_dentry_init_security(struct dentry *dentry, int mode, struct qstr *name,
458 const char **secctx_name, void **secctx,
460 int ll_inode_init_security(struct dentry *dentry, struct inode *inode,
463 int ll_listsecurity(struct inode *inode, char *secctx_name,
464 size_t secctx_name_size);
466 static inline bool obd_connect_has_enc(struct obd_connect_data *data)
468 #ifdef HAVE_LUSTRE_CRYPTO
469 return data->ocd_connect_flags & OBD_CONNECT_FLAGS2 &&
470 data->ocd_connect_flags2 & OBD_CONNECT2_ENCRYPT;
476 static inline void obd_connect_set_enc(struct obd_connect_data *data)
478 #ifdef HAVE_LUSTRE_CRYPTO
479 data->ocd_connect_flags2 |= OBD_CONNECT2_ENCRYPT;
484 * Locking to guarantee consistency of non-atomic updates to long long i_size,
485 * consistency between file size and KMS.
487 * Implemented by ->lli_size_mutex and ->lsm_lock, nested in that order.
490 void ll_inode_size_lock(struct inode *inode);
491 void ll_inode_size_unlock(struct inode *inode);
493 static inline struct ll_inode_info *ll_i2info(struct inode *inode)
495 return container_of(inode, struct ll_inode_info, lli_vfs_inode);
498 static inline struct pcc_inode *ll_i2pcci(struct inode *inode)
500 return ll_i2info(inode)->lli_pcc_inode;
503 /* default to use at least 16M for fast read if possible */
504 #define RA_REMAIN_WINDOW_MIN MiB_TO_PAGES(16UL)
506 /* default read-ahead on a given client mountpoint. */
507 #define SBI_DEFAULT_READ_AHEAD_MAX MiB_TO_PAGES(1024UL)
509 /* default read-ahead for a single file descriptor */
510 #define SBI_DEFAULT_READ_AHEAD_PER_FILE_MAX MiB_TO_PAGES(256UL)
512 /* default read-ahead full files smaller than limit on the second read */
513 #define SBI_DEFAULT_READ_AHEAD_WHOLE_MAX MiB_TO_PAGES(2UL)
515 /* default range pages */
516 #define SBI_DEFAULT_RA_RANGE_PAGES MiB_TO_PAGES(1ULL)
518 /* Min range pages */
519 #define RA_MIN_MMAP_RANGE_PAGES 16UL
524 RA_STAT_DISTANT_READPAGE,
525 RA_STAT_MISS_IN_WINDOW,
526 RA_STAT_FAILED_GRAB_PAGE,
527 RA_STAT_FAILED_MATCH,
532 RA_STAT_MAX_IN_FLIGHT,
533 RA_STAT_WRONG_GRAB_PAGE,
534 RA_STAT_FAILED_REACH_END,
536 RA_STAT_FAILED_FAST_READ,
537 RA_STAT_MMAP_RANGE_READ,
542 atomic_t ra_cur_pages;
543 unsigned long ra_max_pages;
544 unsigned long ra_max_pages_per_file;
545 unsigned long ra_range_pages;
546 unsigned long ra_max_read_ahead_whole_pages;
547 struct workqueue_struct *ll_readahead_wq;
549 * Max number of active works could be triggered
550 * for async readahead.
552 unsigned int ra_async_max_active;
553 /* how many async readahead triggered in flight */
554 atomic_t ra_async_inflight;
555 /* Threshold to control when to trigger async readahead */
556 unsigned long ra_async_pages_per_file_threshold;
559 /* ra_io_arg will be filled in the beginning of ll_readahead with
560 * ras_lock, then the following ll_read_ahead_pages will read RA
561 * pages according to this arg, all the items in this structure are
562 * counted by page index.
565 pgoff_t ria_start_idx; /* start offset of read-ahead*/
566 pgoff_t ria_end_idx; /* end offset of read-ahead*/
567 unsigned long ria_reserved; /* reserved pages for read-ahead */
568 pgoff_t ria_end_idx_min;/* minimum end to cover current read */
569 bool ria_eof; /* reach end of file */
570 /* If stride read pattern is detected, ria_stoff is the byte offset
571 * where stride read is started. Note: for normal read-ahead, the
572 * value here is meaningless, and also it will not be accessed*/
574 /* ria_length and ria_bytes are the length and pages length in the
575 * stride I/O mode. And they will also be used to check whether
576 * it is stride I/O read-ahead in the read-ahead pages*/
581 /* LL_HIST_MAX=32 causes an overflow */
582 #define LL_HIST_MAX 28
583 #define LL_HIST_START 12 /* buckets start at 2^12 = 4k */
584 #define LL_PROCESS_HIST_MAX 10
585 struct per_process_info {
587 struct obd_histogram pp_r_hist;
588 struct obd_histogram pp_w_hist;
591 /* pp_extents[LL_PROCESS_HIST_MAX] will hold the combined process info */
592 struct ll_rw_extents_info {
594 struct per_process_info pp_extents[LL_PROCESS_HIST_MAX + 1];
597 #define LL_OFFSET_HIST_MAX 100
598 struct ll_rw_process_info {
601 loff_t rw_range_start;
603 loff_t rw_last_file_pos;
605 size_t rw_smallest_extent;
606 size_t rw_largest_extent;
607 struct ll_file_data *rw_last_file;
610 enum stats_track_type {
611 STATS_TRACK_ALL = 0, /* track all processes */
612 STATS_TRACK_PID, /* track process with this pid */
613 STATS_TRACK_PPID, /* track processes with this ppid */
614 STATS_TRACK_GID, /* track processes with this gid */
618 /* flags for sbi->ll_flags */
620 LL_SBI_NOLCK, /* DLM locking disabled directio-only */
621 LL_SBI_CHECKSUM, /* checksum each page as it's written */
622 LL_SBI_LOCALFLOCK, /* local flocks instead of fs-wide */
623 LL_SBI_FLOCK, /* flock enabled */
624 LL_SBI_USER_XATTR, /* support user xattr */
625 LL_SBI_LRU_RESIZE, /* lru resize support */
626 LL_SBI_LAZYSTATFS, /* lazystatfs mount option */
627 LL_SBI_32BIT_API, /* generate 32 bit inodes. */
628 LL_SBI_USER_FID2PATH, /* fid2path by unprivileged users */
629 LL_SBI_VERBOSE, /* verbose mount/umount */
630 LL_SBI_ALWAYS_PING, /* ping even if server suppress_pings */
631 LL_SBI_TEST_DUMMY_ENCRYPTION, /* test dummy encryption */
632 LL_SBI_ENCRYPT, /* client side encryption */
633 LL_SBI_FOREIGN_SYMLINK, /* foreign fake-symlink support */
634 LL_SBI_FOREIGN_SYMLINK_UPCALL, /* foreign fake-symlink upcall set */
635 LL_SBI_NUM_MOUNT_OPT,
637 LL_SBI_ACL, /* support ACL */
638 LL_SBI_AGL_ENABLED, /* enable agl */
639 LL_SBI_64BIT_HASH, /* support 64-bits dir hash/offset */
640 LL_SBI_LAYOUT_LOCK, /* layout lock support */
641 LL_SBI_XATTR_CACHE, /* support for xattr cache */
642 LL_SBI_NOROOTSQUASH, /* do not apply root squash */
643 LL_SBI_FAST_READ, /* fast read support */
644 LL_SBI_FILE_SECCTX, /* file security context at create */
645 LL_SBI_TINY_WRITE, /* tiny write support */
646 LL_SBI_FILE_HEAT, /* file heat support */
647 LL_SBI_PARALLEL_DIO, /* parallel (async) O_DIRECT RPCs */
651 int ll_sbi_flags_seq_show(struct seq_file *m, void *v);
653 /* This is embedded into llite super-blocks to keep track of connect
654 * flags (capabilities) supported by all imports given mount is
656 struct lustre_client_ocd {
657 /* This is conjunction of connect_flags across all imports
658 * (LOVs) this mount is connected to. This field is updated by
659 * cl_ocd_update() under ->lco_lock. */
661 struct mutex lco_lock;
662 struct obd_export *lco_md_exp;
663 struct obd_export *lco_dt_exp;
667 /* this protects pglist and ra_info. It isn't safe to
668 * grab from interrupt contexts */
670 spinlock_t ll_pp_extent_lock; /* pp_extent entry*/
671 spinlock_t ll_process_lock; /* ll_rw_process_info */
672 struct obd_uuid ll_sb_uuid;
673 struct obd_export *ll_md_exp;
674 struct obd_export *ll_dt_exp;
675 struct obd_device *ll_md_obd;
676 struct obd_device *ll_dt_obd;
677 struct dentry *ll_debugfs_entry;
678 struct lu_fid ll_root_fid; /* root object fid */
679 struct mnt_namespace *ll_mnt_ns;
681 DECLARE_BITMAP(ll_flags, LL_SBI_NUM_FLAGS); /* enum ll_sbi_flags */
682 unsigned int ll_xattr_cache_enabled:1,
683 ll_xattr_cache_set:1, /* already set to 0/1 */
684 ll_client_common_fill_super_succeeded:1,
687 struct lustre_client_ocd ll_lco;
689 struct lprocfs_stats *ll_stats; /* lprocfs stats counter */
691 /* Used to track "unstable" pages on a client, and maintain a
692 * LRU list of clean pages. An "unstable" page is defined as
693 * any page which is sent to a server as part of a bulk request,
694 * but is uncommitted to stable storage. */
695 struct cl_client_cache *ll_cache;
697 struct lprocfs_stats *ll_ra_stats;
699 struct ll_ra_info ll_ra_info;
700 unsigned int ll_namelen;
701 const struct file_operations *ll_fop;
703 struct lu_site *ll_site;
704 struct cl_device *ll_cl;
707 struct ll_rw_extents_info *ll_rw_extents_info;
708 int ll_extent_process_count;
709 unsigned int ll_offset_process_count;
710 struct ll_rw_process_info *ll_rw_process_info;
711 struct ll_rw_process_info *ll_rw_offset_info;
712 ktime_t ll_process_stats_init;
713 unsigned int ll_rw_offset_entry_count;
714 int ll_stats_track_id;
715 enum stats_track_type ll_stats_track_type;
718 /* metadata stat-ahead */
719 unsigned int ll_sa_running_max;/* max concurrent
720 * statahead instances */
721 unsigned int ll_sa_max; /* max statahead RPCs */
722 atomic_t ll_sa_total; /* statahead thread started
724 atomic_t ll_sa_wrong; /* statahead thread stopped for
726 atomic_t ll_sa_running; /* running statahead thread
728 atomic_t ll_agl_total; /* AGL thread started count */
730 dev_t ll_sdev_orig; /* save s_dev before assign for
733 struct root_squash_info ll_squash;
736 /* st_blksize returned by stat(2), when non-zero */
737 unsigned int ll_stat_blksize;
739 /* maximum relative age of cached statfs results */
740 unsigned int ll_statfs_max_age;
742 struct kset ll_kset; /* sysfs object */
743 struct completion ll_kobj_unregister;
746 unsigned int ll_heat_decay_weight;
747 unsigned int ll_heat_period_second;
749 /* Opens of the same inode before we start requesting open lock */
750 u32 ll_oc_thrsh_count;
752 /* Time in ms between last inode close and next open to be considered
753 * instant back to back and would trigger an open lock request
757 /* Time in ms after last file close that we no longer count prior opens*/
760 /* filesystem fsname */
761 char ll_fsname[LUSTRE_MAXFSNAME + 1];
763 /* Persistent Client Cache */
764 struct pcc_super ll_pcc_super;
766 /* to protect vs updates in all following foreign symlink fields */
767 struct rw_semaphore ll_foreign_symlink_sem;
768 /* foreign symlink path prefix */
769 char *ll_foreign_symlink_prefix;
770 /* full prefix size including leading '\0' */
771 size_t ll_foreign_symlink_prefix_size;
772 /* foreign symlink path upcall */
773 char *ll_foreign_symlink_upcall;
774 /* foreign symlink path upcall infos */
775 struct ll_foreign_symlink_upcall_item *ll_foreign_symlink_upcall_items;
776 /* foreign symlink path upcall nb infos */
777 unsigned int ll_foreign_symlink_upcall_nb_items;
780 #define SBI_DEFAULT_HEAT_DECAY_WEIGHT ((80 * 256 + 50) / 100)
781 #define SBI_DEFAULT_HEAT_PERIOD_SECOND (60)
783 #define SBI_DEFAULT_OPENCACHE_THRESHOLD_COUNT (5)
784 #define SBI_DEFAULT_OPENCACHE_THRESHOLD_MS (100) /* 0.1 second */
785 #define SBI_DEFAULT_OPENCACHE_THRESHOLD_MAX_MS (60000) /* 1 minute */
788 * per file-descriptor read-ahead data.
790 struct ll_readahead_state {
792 /* End byte that read(2) try to read. */
793 loff_t ras_last_read_end_bytes;
795 * number of bytes read after last read-ahead window reset. As window
796 * is reset on each seek, this is effectively a number of consecutive
797 * accesses. Maybe ->ras_accessed_in_window is better name.
799 * XXX nikita: window is also reset (by ras_update()) when Lustre
800 * believes that memory pressure evicts read-ahead pages. In that
801 * case, it probably doesn't make sense to expand window to
802 * PTLRPC_MAX_BRW_PAGES on the third access.
804 loff_t ras_consecutive_bytes;
806 * number of read requests after the last read-ahead window reset
807 * As window is reset on each seek, this is effectively the number
808 * on consecutive read request and is used to trigger read-ahead.
810 unsigned long ras_consecutive_requests;
812 * Parameters of current read-ahead window. Handled by
813 * ras_update(). On the initial access to the file or after a seek,
814 * window is reset to 0. After 3 consecutive accesses, window is
815 * expanded to PTLRPC_MAX_BRW_PAGES. Afterwards, window is enlarged by
816 * PTLRPC_MAX_BRW_PAGES chunks up to ->ra_max_pages.
818 pgoff_t ras_window_start_idx;
819 pgoff_t ras_window_pages;
821 /* Page index where min range read starts */
822 pgoff_t ras_range_min_start_idx;
823 /* Page index where mmap range read ends */
824 pgoff_t ras_range_max_end_idx;
825 /* number of mmap pages where last time detected */
826 pgoff_t ras_last_range_pages;
827 /* number of mmap range requests */
828 pgoff_t ras_range_requests;
831 * Optimal RPC size in pages.
832 * It decides how many pages will be sent for each read-ahead.
834 unsigned long ras_rpc_pages;
836 * Where next read-ahead should start at. This lies within read-ahead
837 * window. Read-ahead window is read in pieces rather than at once
838 * because: 1. lustre limits total number of pages under read-ahead by
839 * ->ra_max_pages (see ll_ra_count_get()), 2. client cannot read pages
840 * not covered by DLM lock.
842 pgoff_t ras_next_readahead_idx;
844 * Total number of ll_file_read requests issued, reads originating
845 * due to mmap are not counted in this total. This value is used to
846 * trigger full file read-ahead after multiple reads to a small file.
848 unsigned long ras_requests;
850 * The following 3 items are used for detecting the stride I/O
852 * In stride I/O mode,
853 * ...............|-----data-----|****gap*****|--------|******|....
854 * offset |-stride_bytes-|-stride_gap-|
855 * ras_stride_offset = offset;
856 * ras_stride_length = stride_bytes + stride_gap;
857 * ras_stride_bytes = stride_bytes;
858 * Note: all these three items are counted by bytes.
860 loff_t ras_stride_offset;
861 loff_t ras_stride_length;
862 loff_t ras_stride_bytes;
864 * number of consecutive stride request count, and it is similar as
865 * ras_consecutive_requests, but used for stride I/O mode.
866 * Note: only more than 2 consecutive stride request are detected,
867 * stride read-ahead will be enable
869 unsigned long ras_consecutive_stride_requests;
870 /* index of the last page that async readahead starts */
871 pgoff_t ras_async_last_readpage_idx;
872 /* whether we should increase readahead window */
873 bool ras_need_increase_window;
874 /* whether ra miss check should be skipped */
875 bool ras_no_miss_check;
878 struct ll_readahead_work {
879 /** File to readahead */
880 struct file *lrw_file;
881 pgoff_t lrw_start_idx;
884 /* async worker to handler read */
885 struct work_struct lrw_readahead_work;
886 char lrw_jobid[LUSTRE_JOBID_SIZE];
889 extern struct kmem_cache *ll_file_data_slab;
890 struct lustre_handle;
891 struct ll_file_data {
892 struct ll_readahead_state fd_ras;
893 struct ll_grouplock fd_grouplock;
897 /* openhandle if lease exists for this file.
898 * Borrow lli->lli_och_mutex to protect assignment */
899 struct obd_client_handle *fd_lease_och;
900 struct obd_client_handle *fd_och;
901 struct file *fd_file;
902 /* Indicate whether need to report failure when close.
903 * true: failure is known, not report again.
904 * false: unknown failure, should report. */
905 bool fd_write_failed;
906 bool ll_lock_no_expand;
907 /* Used by mirrored file to lead IOs to a specific mirror, usually
908 * for mirror resync. 0 means default. */
909 __u32 fd_designated_mirror;
910 /* The layout version when resync starts. Resync I/O should carry this
911 * layout version for verification to OST objects */
912 __u32 fd_layout_version;
913 struct pcc_file fd_pcc_file;
916 void llite_tunables_unregister(void);
917 int llite_tunables_register(void);
919 static inline struct inode *ll_info2i(struct ll_inode_info *lli)
921 return &lli->lli_vfs_inode;
924 __u32 ll_i2suppgid(struct inode *i);
925 void ll_i2gids(__u32 *suppgids, struct inode *i1,struct inode *i2);
927 static inline int ll_need_32bit_api(struct ll_sb_info *sbi)
929 #if BITS_PER_LONG == 32
931 #elif defined(CONFIG_COMPAT)
932 if (unlikely(test_bit(LL_SBI_32BIT_API, sbi->ll_flags)))
935 # ifdef CONFIG_X86_X32
936 /* in_compat_syscall() returns true when called from a kthread
937 * and CONFIG_X86_X32 is enabled, which is wrong. So check
938 * whether the caller comes from a syscall (ie. not a kthread)
939 * before calling in_compat_syscall(). */
940 if (current->flags & PF_KTHREAD)
944 return unlikely(in_compat_syscall());
946 return unlikely(test_bit(LL_SBI_32BIT_API, sbi->ll_flags));
950 static inline bool ll_sbi_has_fast_read(struct ll_sb_info *sbi)
952 return test_bit(LL_SBI_FAST_READ, sbi->ll_flags);
955 static inline bool ll_sbi_has_tiny_write(struct ll_sb_info *sbi)
957 return test_bit(LL_SBI_TINY_WRITE, sbi->ll_flags);
960 static inline bool ll_sbi_has_file_heat(struct ll_sb_info *sbi)
962 return test_bit(LL_SBI_FILE_HEAT, sbi->ll_flags);
965 static inline bool ll_sbi_has_foreign_symlink(struct ll_sb_info *sbi)
967 return test_bit(LL_SBI_FOREIGN_SYMLINK, sbi->ll_flags);
970 static inline bool ll_sbi_has_parallel_dio(struct ll_sb_info *sbi)
972 return test_bit(LL_SBI_PARALLEL_DIO, sbi->ll_flags);
975 void ll_ras_enter(struct file *f, loff_t pos, size_t count);
977 /* llite/lcommon_misc.c */
978 int cl_ocd_update(struct obd_device *host, struct obd_device *watched,
979 enum obd_notify_event ev, void *owner);
980 int cl_get_grouplock(struct cl_object *obj, unsigned long gid, int nonblock,
981 struct ll_grouplock *lg);
982 void cl_put_grouplock(struct ll_grouplock *lg);
984 /* llite/lproc_llite.c */
985 int ll_debugfs_register_super(struct super_block *sb, const char *name);
986 void ll_debugfs_unregister_super(struct super_block *sb);
987 void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, long count);
988 void ll_free_rw_stats_info(struct ll_sb_info *sbi);
992 LPROC_LL_WRITE_BYTES,
1019 LPROC_LL_GETXATTR_HITS,
1021 LPROC_LL_REMOVEXATTR,
1022 LPROC_LL_INODE_PERM,
1024 LPROC_LL_INODE_OCOUNT,
1025 LPROC_LL_INODE_OPCLTM,
1026 LPROC_LL_FILE_OPCODES
1030 enum get_default_layout_type {
1031 GET_DEFAULT_LAYOUT_ROOT = 1,
1034 extern const struct file_operations ll_dir_operations;
1035 extern const struct inode_operations ll_dir_inode_operations;
1036 #ifdef HAVE_DIR_CONTEXT
1037 int ll_dir_read(struct inode *inode, __u64 *pos, struct md_op_data *op_data,
1038 struct dir_context *ctx);
1040 int ll_dir_read(struct inode *inode, __u64 *pos, struct md_op_data *op_data,
1041 void *cookie, filldir_t filldir);
1043 int ll_get_mdt_idx(struct inode *inode);
1044 int ll_get_mdt_idx_by_fid(struct ll_sb_info *sbi, const struct lu_fid *fid);
1045 struct page *ll_get_dir_page(struct inode *dir, struct md_op_data *op_data,
1047 void ll_release_page(struct inode *inode, struct page *page, bool remove);
1048 int quotactl_ioctl(struct super_block *sb, struct if_quotactl *qctl);
1051 extern const struct inode_operations ll_special_inode_operations;
1053 struct inode *ll_iget(struct super_block *sb, ino_t hash,
1054 struct lustre_md *lic);
1055 int ll_test_inode_by_fid(struct inode *inode, void *opaque);
1056 int ll_md_blocking_ast(struct ldlm_lock *, struct ldlm_lock_desc *,
1057 void *data, int flag);
1058 struct dentry *ll_splice_alias(struct inode *inode, struct dentry *de);
1059 int ll_rmdir_entry(struct inode *dir, char *name, int namelen);
1060 void ll_update_times(struct ptlrpc_request *request, struct inode *inode);
1063 int ll_writepage(struct page *page, struct writeback_control *wbc);
1064 int ll_writepages(struct address_space *, struct writeback_control *wbc);
1065 int ll_readpage(struct file *file, struct page *page);
1066 int ll_io_read_page(const struct lu_env *env, struct cl_io *io,
1067 struct cl_page *page, struct file *file);
1068 void ll_readahead_init(struct inode *inode, struct ll_readahead_state *ras);
1069 int vvp_io_write_commit(const struct lu_env *env, struct cl_io *io);
1072 void ll_cl_add(struct inode *inode, const struct lu_env *env, struct cl_io *io,
1073 enum lcc_type type);
1074 void ll_cl_remove(struct inode *inode, const struct lu_env *env);
1075 struct ll_cl_context *ll_cl_find(struct inode *inode);
1077 extern const struct address_space_operations ll_aops;
1080 extern const struct inode_operations ll_file_inode_operations;
1081 const struct file_operations *ll_select_file_operations(struct ll_sb_info *sbi);
1082 extern int ll_have_md_lock(struct inode *inode, __u64 *bits,
1083 enum ldlm_mode l_req_mode);
1084 extern enum ldlm_mode ll_take_md_lock(struct inode *inode, __u64 bits,
1085 struct lustre_handle *lockh, __u64 flags,
1086 enum ldlm_mode mode);
1088 int ll_file_open(struct inode *inode, struct file *file);
1089 int ll_file_release(struct inode *inode, struct file *file);
1090 int ll_release_openhandle(struct dentry *, struct lookup_intent *);
1091 int ll_md_real_close(struct inode *inode, fmode_t fmode);
1092 void ll_track_file_opens(struct inode *inode);
1093 extern void ll_rw_stats_tally(struct ll_sb_info *sbi, pid_t pid,
1094 struct ll_file_data *file, loff_t pos,
1095 size_t count, int rw);
1096 #ifdef HAVE_INODEOPS_ENHANCED_GETATTR
1097 int ll_getattr(const struct path *path, struct kstat *stat,
1098 u32 request_mask, unsigned int flags);
1100 int ll_getattr(struct vfsmount *mnt, struct dentry *de, struct kstat *stat);
1102 int ll_getattr_dentry(struct dentry *de, struct kstat *stat, u32 request_mask,
1103 unsigned int flags, bool foreign);
1104 #ifdef CONFIG_LUSTRE_FS_POSIX_ACL
1105 struct posix_acl *ll_get_acl(struct inode *inode, int type);
1106 int ll_set_acl(struct inode *inode, struct posix_acl *acl, int type);
1107 #else /* !CONFIG_LUSTRE_FS_POSIX_ACL */
1108 #define ll_get_acl NULL
1109 #define ll_set_acl NULL
1110 #endif /* CONFIG_LUSTRE_FS_POSIX_ACL */
1112 static inline int ll_xflags_to_inode_flags(int xflags)
1114 return ((xflags & FS_XFLAG_SYNC) ? S_SYNC : 0) |
1115 ((xflags & FS_XFLAG_NOATIME) ? S_NOATIME : 0) |
1116 ((xflags & FS_XFLAG_APPEND) ? S_APPEND : 0) |
1117 ((xflags & FS_XFLAG_IMMUTABLE) ? S_IMMUTABLE : 0);
1120 static inline int ll_inode_flags_to_xflags(int inode_flags)
1122 return ((inode_flags & S_SYNC) ? FS_XFLAG_SYNC : 0) |
1123 ((inode_flags & S_NOATIME) ? FS_XFLAG_NOATIME : 0) |
1124 ((inode_flags & S_APPEND) ? FS_XFLAG_APPEND : 0) |
1125 ((inode_flags & S_IMMUTABLE) ? FS_XFLAG_IMMUTABLE : 0);
1128 int ll_migrate(struct inode *parent, struct file *file,
1129 struct lmv_user_md *lum, const char *name, __u32 flags);
1130 int ll_get_fid_by_name(struct inode *parent, const char *name,
1131 int namelen, struct lu_fid *fid, struct inode **inode);
1132 int ll_inode_permission(struct inode *inode, int mask);
1133 int ll_ioctl_check_project(struct inode *inode, __u32 xflags, __u32 projid);
1134 int ll_ioctl_fsgetxattr(struct inode *inode, unsigned int cmd,
1136 int ll_ioctl_fssetxattr(struct inode *inode, unsigned int cmd,
1138 int ll_ioctl_project(struct file *file, unsigned int cmd,
1141 int ll_lov_setstripe_ea_info(struct inode *inode, struct dentry *dentry,
1142 __u64 flags, struct lov_user_md *lum,
1144 int ll_lov_getstripe_ea_info(struct inode *inode, const char *filename,
1145 struct lov_mds_md **lmm, int *lmm_size,
1146 struct ptlrpc_request **request);
1147 int ll_dir_setstripe(struct inode *inode, struct lov_user_md *lump,
1149 int ll_dir_getstripe_default(struct inode *inode, void **lmmp,
1150 int *lmm_size, struct ptlrpc_request **request,
1151 struct ptlrpc_request **root_request, u64 valid);
1152 int ll_dir_getstripe(struct inode *inode, void **plmm, int *plmm_size,
1153 struct ptlrpc_request **request, u64 valid);
1154 int ll_fsync(struct file *file, loff_t start, loff_t end, int data);
1155 int ll_merge_attr(const struct lu_env *env, struct inode *inode);
1156 int ll_fid2path(struct inode *inode, void __user *arg);
1157 int ll_data_version(struct inode *inode, __u64 *data_version, int flags);
1158 int ll_hsm_release(struct inode *inode);
1159 int ll_hsm_state_set(struct inode *inode, struct hsm_state_set *hss);
1160 void ll_io_set_mirror(struct cl_io *io, const struct file *file);
1162 /* llite/dcache.c */
1164 extern const struct dentry_operations ll_d_ops;
1166 bool ll_d_setup(struct dentry *de, bool do_put);
1168 static inline bool lld_is_init(struct dentry *dentry)
1170 return ll_d2d(dentry);
1173 #define ll_d_setup(de, do_put) (true)
1174 #define lld_is_init(dentry) (true)
1177 void ll_intent_drop_lock(struct lookup_intent *);
1178 void ll_intent_release(struct lookup_intent *);
1179 void ll_prune_aliases(struct inode *inode);
1180 void ll_lookup_finish_locks(struct lookup_intent *it, struct dentry *dentry);
1181 int ll_revalidate_it_finish(struct ptlrpc_request *request,
1182 struct lookup_intent *it, struct dentry *de);
1184 /* llite/llite_lib.c */
1185 extern const struct super_operations lustre_super_operations;
1187 void ll_lli_init(struct ll_inode_info *lli);
1188 int ll_fill_super(struct super_block *sb);
1189 void ll_put_super(struct super_block *sb);
1190 void ll_kill_super(struct super_block *sb);
1191 struct inode *ll_inode_from_resource_lock(struct ldlm_lock *lock);
1192 void ll_dir_clear_lsm_md(struct inode *inode);
1193 void ll_clear_inode(struct inode *inode);
1194 int volatile_ref_file(const char *volatile_name, int volatile_len,
1195 struct file **ref_file);
1196 int ll_setattr_raw(struct dentry *dentry, struct iattr *attr,
1197 enum op_xvalid xvalid, bool hsm_import);
1198 int ll_setattr(struct dentry *de, struct iattr *attr);
1199 int ll_statfs(struct dentry *de, struct kstatfs *sfs);
1200 int ll_statfs_internal(struct ll_sb_info *sbi, struct obd_statfs *osfs,
1202 int ll_update_inode(struct inode *inode, struct lustre_md *md);
1203 void ll_update_inode_flags(struct inode *inode, unsigned int ext_flags);
1204 void ll_update_dir_depth(struct inode *dir, struct inode *inode);
1205 int ll_read_inode2(struct inode *inode, void *opaque);
1206 void ll_truncate_inode_pages_final(struct inode *inode);
1207 void ll_delete_inode(struct inode *inode);
1208 int ll_iocontrol(struct inode *inode, struct file *file,
1209 unsigned int cmd, unsigned long arg);
1210 int ll_flush_ctx(struct inode *inode);
1211 void ll_umount_begin(struct super_block *sb);
1212 int ll_remount_fs(struct super_block *sb, int *flags, char *data);
1213 int ll_show_options(struct seq_file *seq, struct dentry *dentry);
1214 void ll_dirty_page_discard_warn(struct page *page, int ioret);
1215 int ll_prep_inode(struct inode **inode, struct req_capsule *pill,
1216 struct super_block *sb, struct lookup_intent *it);
1217 int ll_obd_statfs(struct inode *inode, void __user *arg);
1218 int ll_get_max_mdsize(struct ll_sb_info *sbi, int *max_mdsize);
1219 int ll_get_default_mdsize(struct ll_sb_info *sbi, int *default_mdsize);
1220 int ll_set_default_mdsize(struct ll_sb_info *sbi, int default_mdsize);
1222 void ll_unlock_md_op_lsm(struct md_op_data *op_data);
1223 struct md_op_data *ll_prep_md_op_data(struct md_op_data *op_data,
1224 struct inode *i1, struct inode *i2,
1225 const char *name, size_t namelen,
1226 __u32 mode, enum md_op_code opc,
1228 void ll_finish_md_op_data(struct md_op_data *op_data);
1229 int ll_get_obd_name(struct inode *inode, unsigned int cmd, unsigned long arg);
1230 void ll_compute_rootsquash_state(struct ll_sb_info *sbi);
1231 ssize_t ll_copy_user_md(const struct lov_user_md __user *md,
1232 struct lov_user_md **kbuf);
1233 void ll_open_cleanup(struct super_block *sb, struct req_capsule *pill);
1235 void ll_dom_finish_open(struct inode *inode, struct ptlrpc_request *req);
1237 /* Compute expected user md size when passing in a md from user space */
1238 static inline ssize_t ll_lov_user_md_size(const struct lov_user_md *lum)
1240 switch (lum->lmm_magic) {
1241 case LOV_USER_MAGIC_V1:
1242 return sizeof(struct lov_user_md_v1);
1243 case LOV_USER_MAGIC_V3:
1244 return sizeof(struct lov_user_md_v3);
1245 case LOV_USER_MAGIC_SPECIFIC:
1246 if (lum->lmm_stripe_count > LOV_MAX_STRIPE_COUNT)
1249 return lov_user_md_size(lum->lmm_stripe_count,
1250 LOV_USER_MAGIC_SPECIFIC);
1251 case LOV_USER_MAGIC_COMP_V1:
1252 return ((struct lov_comp_md_v1 *)lum)->lcm_size;
1253 case LOV_USER_MAGIC_FOREIGN:
1254 return foreign_size(lum);
1260 /* llite/llite_nfs.c */
1261 extern const struct export_operations lustre_export_operations;
1262 __u32 get_uuid2int(const char *name, int len);
1263 struct inode *search_inode_for_lustre(struct super_block *sb,
1264 const struct lu_fid *fid);
1265 int ll_dir_get_parent_fid(struct inode *dir, struct lu_fid *parent_fid);
1267 /* llite/symlink.c */
1268 extern const struct inode_operations ll_fast_symlink_inode_operations;
1271 * IO arguments for various VFS I/O interfaces.
1273 struct vvp_io_args {
1274 /** normal/sendfile/splice */
1277 struct kiocb *via_iocb;
1278 struct iov_iter *via_iter;
1288 struct ll_cl_context {
1289 struct list_head lcc_list;
1291 const struct lu_env *lcc_env;
1292 struct cl_io *lcc_io;
1293 struct cl_page *lcc_page;
1294 enum lcc_type lcc_type;
1296 * Get encryption context operation in progress,
1297 * allow getxattr of LL_XATTR_NAME_ENCRYPTION_CONTEXT xattr
1299 unsigned lcc_getencctx:1;
1302 struct ll_thread_info {
1303 struct vvp_io_args lti_args;
1304 struct ra_io_arg lti_ria;
1305 struct ll_cl_context lti_io_ctx;
1308 extern struct lu_context_key ll_thread_key;
1310 static inline struct ll_thread_info *ll_env_info(const struct lu_env *env)
1312 struct ll_thread_info *lti;
1314 lti = lu_context_key_get(&env->le_ctx, &ll_thread_key);
1315 LASSERT(lti != NULL);
1320 static inline struct vvp_io_args *ll_env_args(const struct lu_env *env)
1322 return &ll_env_info(env)->lti_args;
1325 void ll_io_init(struct cl_io *io, struct file *file, enum cl_io_type iot,
1326 struct vvp_io_args *args);
1328 /* llite/llite_mmap.c */
1330 int ll_file_mmap(struct file * file, struct vm_area_struct * vma);
1331 void policy_from_vma(union ldlm_policy_data *policy, struct vm_area_struct *vma,
1332 unsigned long addr, size_t count);
1333 struct vm_area_struct *our_vma(struct mm_struct *mm, unsigned long addr,
1336 #define ll_s2sbi(sb) (s2lsi(sb)->lsi_llsbi)
1338 /* don't need an addref as the sb_info should be holding one */
1339 static inline struct obd_export *ll_s2dtexp(struct super_block *sb)
1341 return ll_s2sbi(sb)->ll_dt_exp;
1344 /* don't need an addref as the sb_info should be holding one */
1345 static inline struct obd_export *ll_s2mdexp(struct super_block *sb)
1347 return ll_s2sbi(sb)->ll_md_exp;
1350 static inline struct client_obd *sbi2mdc(struct ll_sb_info *sbi)
1352 struct obd_device *obd = sbi->ll_md_exp->exp_obd;
1358 // FIXME: replace the name of this with LL_SB to conform to kernel stuff
1359 static inline struct ll_sb_info *ll_i2sbi(struct inode *inode)
1361 return ll_s2sbi(inode->i_sb);
1364 static inline struct obd_export *ll_i2dtexp(struct inode *inode)
1366 return ll_s2dtexp(inode->i_sb);
1369 static inline struct obd_export *ll_i2mdexp(struct inode *inode)
1371 return ll_s2mdexp(inode->i_sb);
1374 static inline struct lu_fid *ll_inode2fid(struct inode *inode)
1378 LASSERT(inode != NULL);
1379 fid = &ll_i2info(inode)->lli_fid;
1384 static inline bool ll_dir_striped(struct inode *inode)
1387 return S_ISDIR(inode->i_mode) &&
1388 lmv_dir_striped(ll_i2info(inode)->lli_lsm_md);
1391 static inline loff_t ll_file_maxbytes(struct inode *inode)
1393 struct cl_object *obj = ll_i2info(inode)->lli_clob;
1396 return MAX_LFS_FILESIZE;
1398 return min_t(loff_t, cl_object_maxbytes(obj), MAX_LFS_FILESIZE);
1402 extern const struct xattr_handler *ll_xattr_handlers[];
1404 #define XATTR_USER_T 1
1405 #define XATTR_TRUSTED_T 2
1406 #define XATTR_SECURITY_T 3
1407 #define XATTR_ACL_ACCESS_T 4
1408 #define XATTR_ACL_DEFAULT_T 5
1409 #define XATTR_LUSTRE_T 6
1410 #define XATTR_OTHER_T 7
1412 ssize_t ll_listxattr(struct dentry *dentry, char *buffer, size_t size);
1413 int ll_xattr_list(struct inode *inode, const char *name, int type,
1414 void *buffer, size_t size, u64 valid);
1415 const struct xattr_handler *get_xattr_type(const char *name);
1418 * Common IO arguments for various VFS I/O interfaces.
1420 int cl_sb_init(struct super_block *sb);
1421 int cl_sb_fini(struct super_block *sb);
1423 enum ras_update_flags {
1427 void ll_ra_count_put(struct ll_sb_info *sbi, unsigned long len);
1428 void ll_ra_stats_inc(struct inode *inode, enum ra_stat which);
1432 #define LL_SA_RPC_MIN 2
1433 #define LL_SA_RPC_DEF 32
1434 #define LL_SA_RPC_MAX 512
1436 /* XXX: If want to support more concurrent statahead instances,
1437 * please consider to decentralize the RPC lists attached
1438 * on related import, such as imp_{sending,delayed}_list.
1440 #define LL_SA_RUNNING_MAX 256
1441 #define LL_SA_RUNNING_DEF 16
1443 #define LL_SA_CACHE_BIT 5
1444 #define LL_SA_CACHE_SIZE (1 << LL_SA_CACHE_BIT)
1445 #define LL_SA_CACHE_MASK (LL_SA_CACHE_SIZE - 1)
1447 /* per inode struct, for dir only */
1448 struct ll_statahead_info {
1449 struct dentry *sai_dentry;
1450 atomic_t sai_refcount; /* when access this struct, hold
1452 unsigned int sai_max; /* max ahead of lookup */
1453 __u64 sai_sent; /* stat requests sent count */
1454 __u64 sai_replied; /* stat requests which received
1456 __u64 sai_index; /* index of statahead entry */
1457 __u64 sai_index_wait; /* index of entry which is the
1458 * caller is waiting for */
1459 __u64 sai_hit; /* hit count */
1460 __u64 sai_miss; /* miss count:
1461 * for "ls -al" case, includes
1462 * hidden dentry miss;
1463 * for "ls -l" case, it does not
1464 * include hidden dentry miss.
1465 * "sai_miss_hidden" is used for
1468 unsigned int sai_consecutive_miss; /* consecutive miss */
1469 unsigned int sai_miss_hidden;/* "ls -al", but first dentry
1470 * is not a hidden one */
1471 unsigned int sai_skip_hidden;/* skipped hidden dentry count
1473 unsigned int sai_ls_all:1, /* "ls -al", do stat-ahead for
1475 sai_in_readpage:1;/* statahead is in readdir()*/
1476 wait_queue_head_t sai_waitq; /* stat-ahead wait queue */
1477 struct task_struct *sai_task; /* stat-ahead thread */
1478 struct task_struct *sai_agl_task; /* AGL thread */
1479 struct list_head sai_interim_entries; /* entries which got async
1480 * stat reply, but not
1482 struct list_head sai_entries; /* completed entries */
1483 struct list_head sai_agls; /* AGLs to be sent */
1484 struct list_head sai_cache[LL_SA_CACHE_SIZE];
1485 spinlock_t sai_cache_lock[LL_SA_CACHE_SIZE];
1486 atomic_t sai_cache_count; /* entry count in cache */
1489 int ll_revalidate_statahead(struct inode *dir, struct dentry **dentry,
1491 int ll_start_statahead(struct inode *dir, struct dentry *dentry, bool agl);
1492 void ll_authorize_statahead(struct inode *dir, void *key);
1493 void ll_deauthorize_statahead(struct inode *dir, void *key);
1496 blkcnt_t dirty_cnt(struct inode *inode);
1498 int cl_glimpse_size0(struct inode *inode, int agl);
1499 int cl_glimpse_lock(const struct lu_env *env, struct cl_io *io,
1500 struct inode *inode, struct cl_object *clob, int agl);
1502 static inline int cl_glimpse_size(struct inode *inode)
1504 return cl_glimpse_size0(inode, 0);
1507 /* AGL is 'asychronous glimpse lock', which is a speculative lock taken as
1508 * part of statahead */
1509 static inline int cl_agl(struct inode *inode)
1511 return cl_glimpse_size0(inode, 1);
1514 int ll_file_lock_ahead(struct file *file, struct llapi_lu_ladvise *ladvise);
1516 int cl_io_get(struct inode *inode, struct lu_env **envout,
1517 struct cl_io **ioout, __u16 *refcheck);
1519 static inline int ll_glimpse_size(struct inode *inode)
1521 struct ll_inode_info *lli = ll_i2info(inode);
1524 down_read(&lli->lli_glimpse_sem);
1525 rc = cl_glimpse_size(inode);
1526 lli->lli_glimpse_time = ktime_get();
1527 up_read(&lli->lli_glimpse_sem);
1531 /* dentry may statahead when statahead is enabled and current process has opened
1532 * parent directory, and this dentry hasn't accessed statahead cache before */
1534 dentry_may_statahead(struct inode *dir, struct dentry *dentry)
1536 struct ll_inode_info *lli;
1537 struct ll_dentry_data *ldd;
1539 if (ll_i2sbi(dir)->ll_sa_max == 0)
1542 lli = ll_i2info(dir);
1544 /* statahead is not allowed for this dir, there may be three causes:
1545 * 1. dir is not opened.
1546 * 2. statahead hit ratio is too low.
1547 * 3. previous stat started statahead thread failed. */
1548 if (!lli->lli_sa_enabled)
1551 /* not the same process, don't statahead */
1552 if (lli->lli_opendir_pid != current->pid)
1556 * When stating a dentry, kernel may trigger 'revalidate' or 'lookup'
1557 * multiple times, eg. for 'getattr', 'getxattr' and etc.
1558 * For patchless client, lookup intent is not accurate, which may
1559 * misguide statahead. For example:
1560 * The 'revalidate' call for 'getattr' and 'getxattr' of a dentry will
1561 * have the same intent -- IT_GETATTR, while one dentry should access
1562 * statahead cache once, otherwise statahead windows is messed up.
1563 * The solution is as following:
1564 * Assign 'lld_sa_generation' with 'lli_sa_generation' when a dentry
1565 * IT_GETATTR for the first time, and subsequent IT_GETATTR will
1566 * bypass interacting with statahead cache by checking
1567 * 'lld_sa_generation == lli->lli_sa_generation'.
1569 ldd = ll_d2d(dentry);
1570 if (ldd != NULL && lli->lli_sa_generation &&
1571 ldd->lld_sa_generation == lli->lli_sa_generation)
1577 int cl_sync_file_range(struct inode *inode, loff_t start, loff_t end,
1578 enum cl_fsync_mode mode, int ignore_layout);
1580 static inline int ll_file_nolock(const struct file *file)
1582 struct ll_file_data *fd = file->private_data;
1583 struct inode *inode = file_inode((struct file *)file);
1585 LASSERT(fd != NULL);
1586 return ((fd->fd_flags & LL_FILE_IGNORE_LOCK) ||
1587 test_bit(LL_SBI_NOLCK, ll_i2sbi(inode)->ll_flags));
1590 static inline void ll_set_lock_data(struct obd_export *exp, struct inode *inode,
1591 struct lookup_intent *it, __u64 *bits)
1593 if (!it->it_lock_set) {
1594 struct lustre_handle handle;
1596 /* If this inode is a remote object, it will get two
1597 * separate locks in different namespaces, Master MDT,
1598 * where the name entry is, will grant LOOKUP lock,
1599 * remote MDT, where the object is, will grant
1600 * UPDATE|PERM lock. The inode will be attched to both
1601 * LOOKUP and PERM locks, so revoking either locks will
1602 * case the dcache being cleared */
1603 if (it->it_remote_lock_mode) {
1604 handle.cookie = it->it_remote_lock_handle;
1605 CDEBUG(D_DLMTRACE, "setting l_data to inode "DFID
1606 "(%p) for remote lock %#llx\n",
1607 PFID(ll_inode2fid(inode)), inode,
1609 md_set_lock_data(exp, &handle, inode, NULL);
1612 handle.cookie = it->it_lock_handle;
1614 CDEBUG(D_DLMTRACE, "setting l_data to inode "DFID"(%p)"
1615 " for lock %#llx\n",
1616 PFID(ll_inode2fid(inode)), inode, handle.cookie);
1618 md_set_lock_data(exp, &handle, inode, &it->it_lock_bits);
1619 it->it_lock_set = 1;
1623 *bits = it->it_lock_bits;
1626 static inline int d_lustre_invalid(const struct dentry *dentry)
1628 return !ll_d2d(dentry) || ll_d2d(dentry)->lld_invalid;
1632 * Mark dentry INVALID, if dentry refcount is zero (this is normally case for
1633 * ll_md_blocking_ast), it will be pruned by ll_prune_aliases() and
1634 * ll_prune_negative_children(); otherwise dput() of the last refcount will
1635 * unhash this dentry and kill it.
1637 static inline void d_lustre_invalidate(struct dentry *dentry)
1639 CDEBUG(D_DENTRY, "invalidate dentry %pd (%p) parent %p inode %p refc %d\n",
1641 dentry->d_parent, dentry->d_inode, ll_d_count(dentry));
1643 spin_lock(&dentry->d_lock);
1644 if (lld_is_init(dentry))
1645 ll_d2d(dentry)->lld_invalid = 1;
1646 spin_unlock(&dentry->d_lock);
1649 static inline void d_lustre_revalidate(struct dentry *dentry)
1651 spin_lock(&dentry->d_lock);
1652 LASSERT(ll_d2d(dentry));
1653 ll_d2d(dentry)->lld_invalid = 0;
1654 spin_unlock(&dentry->d_lock);
1657 static inline dev_t ll_compat_encode_dev(dev_t dev)
1659 /* The compat_sys_*stat*() syscalls will fail unless the
1660 * device majors and minors are both less than 256. Note that
1661 * the value returned here will be passed through
1662 * old_encode_dev() in cp_compat_stat(). And so we are not
1663 * trying to return a valid compat (u16) device number, just
1664 * one that will pass the old_valid_dev() check. */
1666 return MKDEV(MAJOR(dev) & 0xff, MINOR(dev) & 0xff);
1669 int ll_layout_conf(struct inode *inode, const struct cl_object_conf *conf);
1670 int ll_layout_refresh(struct inode *inode, __u32 *gen);
1671 int ll_layout_restore(struct inode *inode, loff_t start, __u64 length);
1672 int ll_layout_write_intent(struct inode *inode, enum layout_intent_opc opc,
1673 struct lu_extent *ext);
1675 int ll_xattr_init(void);
1676 void ll_xattr_fini(void);
1678 int ll_page_sync_io(const struct lu_env *env, struct cl_io *io,
1679 struct cl_page *page, enum cl_req_type crt);
1681 int ll_getparent(struct file *file, struct getparent __user *arg);
1684 int cl_setattr_ost(struct cl_object *obj, const struct iattr *attr,
1685 enum op_xvalid xvalid, unsigned int attr_flags);
1687 extern struct lu_env *cl_inode_fini_env;
1688 extern __u16 cl_inode_fini_refcheck;
1690 int cl_file_inode_init(struct inode *inode, struct lustre_md *md);
1691 void cl_inode_fini(struct inode *inode);
1693 u64 cl_fid_build_ino(const struct lu_fid *fid, int api32);
1694 u32 cl_fid_build_gen(const struct lu_fid *fid);
1696 static inline struct pcc_super *ll_i2pccs(struct inode *inode)
1698 return &ll_i2sbi(inode)->ll_pcc_super;
1701 static inline struct pcc_super *ll_info2pccs(struct ll_inode_info *lli)
1703 return ll_i2pccs(ll_info2i(lli));
1707 /* The digested form is made of a FID (16 bytes) followed by the second-to-last
1708 * ciphertext block (16 bytes), so a total length of 32 bytes.
1709 * That way, llcrypt does not compute a digested form of this digest.
1711 struct ll_digest_filename {
1712 struct lu_fid ldf_fid;
1713 char ldf_excerpt[LLCRYPT_FNAME_DIGEST_SIZE];
1716 int ll_setup_filename(struct inode *dir, const struct qstr *iname,
1717 int lookup, struct llcrypt_name *fname,
1718 struct lu_fid *fid);
1719 int ll_fname_disk_to_usr(struct inode *inode,
1720 u32 hash, u32 minor_hash,
1721 struct llcrypt_str *iname, struct llcrypt_str *oname,
1722 struct lu_fid *fid);
1723 int ll_revalidate_d_crypto(struct dentry *dentry, unsigned int flags);
1724 int ll_file_open_encrypt(struct inode *inode, struct file *filp);
1725 #ifdef HAVE_LUSTRE_CRYPTO
1726 extern const struct llcrypt_operations lustre_cryptops;
1729 /* llite/llite_foreign.c */
1730 int ll_manage_foreign(struct inode *inode, struct lustre_md *lmd);
1731 bool ll_foreign_is_openable(struct dentry *dentry, unsigned int flags);
1732 bool ll_foreign_is_removable(struct dentry *dentry, bool unset);
1734 #endif /* LLITE_INTERNAL_H */