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/
30 * Lustre is a trademark of Sun Microsystems, Inc.
33 #ifndef LLITE_INTERNAL_H
34 #define LLITE_INTERNAL_H
36 #include <lustre_disk.h> /* for s2sbi */
37 #include <lustre_eacl.h>
38 #include <lustre_linkea.h>
40 /* for struct cl_lock_descr and struct cl_io */
41 #include <cl_object.h>
42 #include <lustre_lmv.h>
43 #include <lustre_mdc.h>
44 #include <lustre_intent.h>
45 #include <linux/compat.h>
46 #include <linux/aio.h>
47 #include <lustre_compat.h>
48 #include <lustre_crypto.h>
49 #include <range_lock.h>
51 #include "vvp_internal.h"
53 #include "foreign_symlink.h"
59 #ifndef HAVE_VM_FAULT_RETRY
60 #define VM_FAULT_RETRY 0
63 /* Kernel 3.1 kills LOOKUP_CONTINUE, LOOKUP_PARENT is equivalent to it.
64 * seem kernel commit 49084c3bb2055c401f3493c13edae14d49128ca0 */
65 #ifndef LOOKUP_CONTINUE
66 #define LOOKUP_CONTINUE LOOKUP_PARENT
69 /** Only used on client-side for indicating the tail of dir hash/offset. */
70 #define LL_DIR_END_OFF 0x7fffffffffffffffULL
71 #define LL_DIR_END_OFF_32BIT 0x7fffffffUL
73 /* 4UL * 1024 * 1024 */
74 #define LL_MAX_BLKSIZE_BITS 22
76 #define LL_IT2STR(it) ((it) ? ldlm_it2str((it)->it_op) : "0")
78 #define TIMES_SET_FLAGS (ATTR_MTIME_SET | ATTR_ATIME_SET | ATTR_TIMES_SET)
80 struct ll_dentry_data {
81 unsigned int lld_sa_generation;
82 unsigned int lld_invalid:1;
83 unsigned int lld_nfs_dentry:1;
84 struct rcu_head lld_rcu_head;
87 #define ll_d2d(de) ((struct ll_dentry_data*)((de)->d_fsdata))
89 #define LLI_INODE_MAGIC 0x111d0de5
90 #define LLI_INODE_DEAD 0xdeadd00d
92 struct ll_getname_data {
93 #ifdef HAVE_DIR_CONTEXT
94 struct dir_context ctx;
96 char *lgd_name; /* points to a buffer with NAME_MAX+1 size */
97 struct lu_fid lgd_fid; /* target fid we are looking for */
98 int lgd_found; /* inode matched? */
101 struct ll_grouplock {
102 struct lu_env *lg_env;
104 struct cl_lock *lg_lock;
105 unsigned long lg_gid;
108 /* See comment on trunc_sem_down_read_nowait */
109 struct ll_trunc_sem {
110 /* when positive, this is a count of readers, when -1, it indicates
111 * the semaphore is held for write, and 0 is unlocked
113 atomic_t ll_trunc_readers;
114 /* this tracks a count of waiting writers */
115 atomic_t ll_trunc_waiters;
118 struct ll_inode_info {
119 __u32 lli_inode_magic;
122 volatile unsigned long lli_flags;
123 struct posix_acl *lli_posix_acl;
125 /* identifying fields for both metadata and data stacks. */
126 struct lu_fid lli_fid;
127 /* master inode fid for stripe directory */
128 struct lu_fid lli_pfid;
130 /* We need all three because every inode may be opened in different
132 struct obd_client_handle *lli_mds_read_och;
133 struct obd_client_handle *lli_mds_write_och;
134 struct obd_client_handle *lli_mds_exec_och;
135 __u64 lli_open_fd_read_count;
136 __u64 lli_open_fd_write_count;
137 __u64 lli_open_fd_exec_count;
138 /* Protects access to och pointers and their usage counters */
139 struct mutex lli_och_mutex;
141 struct inode lli_vfs_inode;
143 /* the most recent timestamps obtained from mds */
148 spinlock_t lli_agl_lock;
150 /* Try to make the d::member and f::member are aligned. Before using
151 * these members, make clear whether it is directory or not. */
155 /* metadata statahead */
156 /* since parent-child threads can share the same @file
157 * struct, "opendir_key" is the token when dir close for
158 * case of parent exit before child -- it is me should
159 * cleanup the dir readahead. */
160 void *lli_opendir_key;
161 struct ll_statahead_info *lli_sai;
162 /* protect statahead stuff. */
163 spinlock_t lli_sa_lock;
164 /* "opendir_pid" is the token when lookup/revalid
165 * -- I am the owner of dir statahead. */
166 pid_t lli_opendir_pid;
167 /* stat will try to access statahead entries or start
168 * statahead if this flag is set, and this flag will be
169 * set upon dir open, and cleared when dir is closed,
170 * statahead hit ratio is too low, or start statahead
172 unsigned int lli_sa_enabled:1;
173 /* generation for statahead */
174 unsigned int lli_sa_generation;
175 /* rw lock protects lli_lsm_md */
176 struct rw_semaphore lli_lsm_sem;
177 /* directory stripe information */
178 struct lmv_stripe_md *lli_lsm_md;
179 /* directory default LMV */
180 struct lmv_stripe_md *lli_default_lsm_md;
183 /* for non-directory */
185 struct mutex lli_size_mutex;
186 char *lli_symlink_name;
187 struct ll_trunc_sem lli_trunc_sem;
188 struct range_lock_tree lli_write_tree;
189 struct mutex lli_setattr_mutex;
191 struct rw_semaphore lli_glimpse_sem;
192 ktime_t lli_glimpse_time;
193 struct list_head lli_agl_list;
196 /* for writepage() only to communicate to fsync */
199 /* protect the file heat fields */
200 spinlock_t lli_heat_lock;
201 __u32 lli_heat_flags;
202 struct obd_heat_instance lli_heat_instances[OBD_HEAT_COUNT];
205 * Whenever a process try to read/write the file, the
206 * jobid of the process will be saved here, and it'll
207 * be packed into the write PRC when flush later.
209 * So the read/write statistics for jobid will not be
210 * accurate if the file is shared by different jobs.
212 char lli_jobid[LUSTRE_JOBID_SIZE];
214 struct mutex lli_pcc_lock;
215 enum lu_pcc_state_flags lli_pcc_state;
217 * @lli_pcc_generation saves the gobal PCC generation
218 * when the file was successfully attached into PCC.
219 * The flags of the PCC dataset are saved in
221 * The gobal PCC generation will be increased when add
222 * or delete a PCC backend, or change the configuration
223 * parameters for PCC.
224 * If @lli_pcc_generation is same as the gobal PCC
225 * generation, we can use the saved flags of the PCC
226 * dataset to determine whether need to try auto attach
229 __u64 lli_pcc_generation;
230 enum pcc_dataset_flags lli_pcc_dsflags;
231 struct pcc_inode *lli_pcc_inode;
233 struct mutex lli_group_mutex;
234 __u64 lli_group_users;
235 unsigned long lli_group_gid;
237 __u64 lli_attr_valid;
239 __u64 lli_lazyblocks;
243 /* XXX: For following frequent used members, although they maybe special
244 * used for non-directory object, it is some time-wasting to check
245 * whether the object is directory or not before using them. On the
246 * other hand, currently, sizeof(f) > sizeof(d), it cannot reduce
247 * the "ll_inode_info" size even if moving those members into u.f.
248 * So keep them out side.
250 * In the future, if more members are added only for directory,
251 * some of the following members can be moved into u.f.
253 struct cl_object *lli_clob;
255 /* mutex to request for layout lock exclusively. */
256 struct mutex lli_layout_mutex;
257 /* Layout version, protected by lli_layout_lock */
258 __u32 lli_layout_gen;
259 spinlock_t lli_layout_lock;
261 __u32 lli_projid; /* project id */
263 struct rw_semaphore lli_xattrs_list_rwsem;
264 struct mutex lli_xattrs_enq_lock;
265 struct list_head lli_xattrs; /* ll_xattr_entry->xe_list */
268 static inline void ll_trunc_sem_init(struct ll_trunc_sem *sem)
270 atomic_set(&sem->ll_trunc_readers, 0);
271 atomic_set(&sem->ll_trunc_waiters, 0);
274 /* This version of down read ignores waiting writers, meaning if the semaphore
275 * is already held for read, this down_read will 'join' that reader and also
276 * take the semaphore.
278 * This lets us avoid an unusual deadlock.
280 * We must take lli_trunc_sem in read mode on entry in to various i/o paths
281 * in Lustre, in order to exclude truncates. Some of these paths then need to
282 * take the mmap_lock, while still holding the trunc_sem. The problem is that
283 * page faults hold the mmap_lock when calling in to Lustre, and then must also
284 * take the trunc_sem to exclude truncate.
286 * This means the locking order for trunc_sem and mmap_lock is sometimes AB,
287 * sometimes BA. This is almost OK because in both cases, we take the trunc
288 * sem for read, so it doesn't block.
290 * However, if a write mode user (truncate, a setattr op) arrives in the
291 * middle of this, the second reader on the truncate_sem will wait behind that
294 * So we have, on our truncate sem, in order (where 'reader' and 'writer' refer
295 * to the mode in which they take the semaphore):
296 * reader (holding mmap_lock, needs truncate_sem)
298 * reader (holding truncate sem, waiting for mmap_lock)
300 * And so the readers deadlock.
302 * The solution is this modified semaphore, where this down_read ignores
303 * waiting write operations, and all waiters are woken up at once, so readers
304 * using down_read_nowait cannot get stuck behind waiting writers, regardless
305 * of the order they arrived in.
307 * down_read_nowait is only used in the page fault case, where we already hold
308 * the mmap_lock. This is because otherwise repeated read and write operations
309 * (which take the truncate sem) could prevent a truncate from ever starting.
310 * This could still happen with page faults, but without an even more complex
311 * mechanism, this is unavoidable.
315 static inline void trunc_sem_down_read_nowait(struct ll_trunc_sem *sem)
317 wait_var_event(&sem->ll_trunc_readers,
318 atomic_inc_unless_negative(&sem->ll_trunc_readers));
321 static inline void trunc_sem_down_read(struct ll_trunc_sem *sem)
323 wait_var_event(&sem->ll_trunc_readers,
324 atomic_read(&sem->ll_trunc_waiters) == 0 &&
325 atomic_inc_unless_negative(&sem->ll_trunc_readers));
328 static inline void trunc_sem_up_read(struct ll_trunc_sem *sem)
330 if (atomic_dec_return(&sem->ll_trunc_readers) == 0 &&
331 atomic_read(&sem->ll_trunc_waiters))
332 wake_up_var(&sem->ll_trunc_readers);
335 static inline void trunc_sem_down_write(struct ll_trunc_sem *sem)
337 atomic_inc(&sem->ll_trunc_waiters);
338 wait_var_event(&sem->ll_trunc_readers,
339 atomic_cmpxchg(&sem->ll_trunc_readers, 0, -1) == 0);
340 atomic_dec(&sem->ll_trunc_waiters);
343 static inline void trunc_sem_up_write(struct ll_trunc_sem *sem)
345 atomic_set(&sem->ll_trunc_readers, 0);
346 wake_up_var(&sem->ll_trunc_readers);
349 #ifdef CONFIG_LUSTRE_FS_POSIX_ACL
350 static inline void lli_clear_acl(struct ll_inode_info *lli)
352 if (lli->lli_posix_acl) {
353 posix_acl_release(lli->lli_posix_acl);
354 lli->lli_posix_acl = NULL;
358 static inline void lli_replace_acl(struct ll_inode_info *lli,
359 struct lustre_md *md)
361 spin_lock(&lli->lli_lock);
362 if (lli->lli_posix_acl)
363 posix_acl_release(lli->lli_posix_acl);
364 lli->lli_posix_acl = md->posix_acl;
365 spin_unlock(&lli->lli_lock);
368 static inline void lli_clear_acl(struct ll_inode_info *lli)
372 static inline void lli_replace_acl(struct ll_inode_info *lli,
373 struct lustre_md *md)
378 static inline __u32 ll_layout_version_get(struct ll_inode_info *lli)
382 spin_lock(&lli->lli_layout_lock);
383 gen = lli->lli_layout_gen;
384 spin_unlock(&lli->lli_layout_lock);
389 static inline void ll_layout_version_set(struct ll_inode_info *lli, __u32 gen)
391 spin_lock(&lli->lli_layout_lock);
392 lli->lli_layout_gen = gen;
393 spin_unlock(&lli->lli_layout_lock);
397 /* File data is modified. */
398 LLIF_DATA_MODIFIED = 0,
399 /* File is being restored */
400 LLIF_FILE_RESTORING = 1,
401 /* Xattr cache is attached to the file */
402 LLIF_XATTR_CACHE = 2,
403 /* Project inherit */
404 LLIF_PROJECT_INHERIT = 3,
405 /* update atime from MDS even if it's older than local inode atime. */
406 LLIF_UPDATE_ATIME = 4,
407 /* foreign file/dir can be unlinked unconditionnaly */
408 LLIF_FOREIGN_REMOVABLE = 5,
412 static inline void ll_file_set_flag(struct ll_inode_info *lli,
413 enum ll_file_flags flag)
415 set_bit(flag, &lli->lli_flags);
418 static inline void ll_file_clear_flag(struct ll_inode_info *lli,
419 enum ll_file_flags flag)
421 clear_bit(flag, &lli->lli_flags);
424 static inline bool ll_file_test_flag(struct ll_inode_info *lli,
425 enum ll_file_flags flag)
427 return test_bit(flag, &lli->lli_flags);
430 static inline bool ll_file_test_and_clear_flag(struct ll_inode_info *lli,
431 enum ll_file_flags flag)
433 return test_and_clear_bit(flag, &lli->lli_flags);
436 int ll_xattr_cache_destroy(struct inode *inode);
438 int ll_xattr_cache_get(struct inode *inode,
444 int ll_xattr_cache_insert(struct inode *inode,
449 static inline bool obd_connect_has_secctx(struct obd_connect_data *data)
451 #ifdef CONFIG_SECURITY
452 return data->ocd_connect_flags & OBD_CONNECT_FLAGS2 &&
453 data->ocd_connect_flags2 & OBD_CONNECT2_FILE_SECCTX;
459 static inline void obd_connect_set_secctx(struct obd_connect_data *data)
461 #ifdef CONFIG_SECURITY
462 data->ocd_connect_flags2 |= OBD_CONNECT2_FILE_SECCTX;
466 int ll_dentry_init_security(struct dentry *dentry, int mode, struct qstr *name,
467 const char **secctx_name, void **secctx,
469 int ll_inode_init_security(struct dentry *dentry, struct inode *inode,
472 int ll_listsecurity(struct inode *inode, char *secctx_name,
473 size_t secctx_name_size);
475 static inline bool obd_connect_has_enc(struct obd_connect_data *data)
477 #ifdef HAVE_LUSTRE_CRYPTO
478 return data->ocd_connect_flags & OBD_CONNECT_FLAGS2 &&
479 data->ocd_connect_flags2 & OBD_CONNECT2_ENCRYPT;
485 static inline void obd_connect_set_enc(struct obd_connect_data *data)
487 #ifdef HAVE_LUSTRE_CRYPTO
488 data->ocd_connect_flags2 |= OBD_CONNECT2_ENCRYPT;
493 * Locking to guarantee consistency of non-atomic updates to long long i_size,
494 * consistency between file size and KMS.
496 * Implemented by ->lli_size_mutex and ->lsm_lock, nested in that order.
499 void ll_inode_size_lock(struct inode *inode);
500 void ll_inode_size_unlock(struct inode *inode);
502 static inline struct ll_inode_info *ll_i2info(struct inode *inode)
504 return container_of(inode, struct ll_inode_info, lli_vfs_inode);
507 static inline struct pcc_inode *ll_i2pcci(struct inode *inode)
509 return ll_i2info(inode)->lli_pcc_inode;
512 /* default to use at least 16M for fast read if possible */
513 #define RA_REMAIN_WINDOW_MIN MiB_TO_PAGES(16UL)
515 /* default read-ahead on a given client mountpoint. */
516 #define SBI_DEFAULT_READ_AHEAD_MAX MiB_TO_PAGES(1024UL)
518 /* default read-ahead for a single file descriptor */
519 #define SBI_DEFAULT_READ_AHEAD_PER_FILE_MAX MiB_TO_PAGES(256UL)
521 /* default read-ahead full files smaller than limit on the second read */
522 #define SBI_DEFAULT_READ_AHEAD_WHOLE_MAX MiB_TO_PAGES(2UL)
524 /* default range pages */
525 #define SBI_DEFAULT_RA_RANGE_PAGES MiB_TO_PAGES(1ULL)
527 /* Min range pages */
528 #define RA_MIN_MMAP_RANGE_PAGES 16UL
533 RA_STAT_DISTANT_READPAGE,
534 RA_STAT_MISS_IN_WINDOW,
535 RA_STAT_FAILED_GRAB_PAGE,
536 RA_STAT_FAILED_MATCH,
541 RA_STAT_MAX_IN_FLIGHT,
542 RA_STAT_WRONG_GRAB_PAGE,
543 RA_STAT_FAILED_REACH_END,
545 RA_STAT_FAILED_FAST_READ,
546 RA_STAT_MMAP_RANGE_READ,
551 atomic_t ra_cur_pages;
552 unsigned long ra_max_pages;
553 unsigned long ra_max_pages_per_file;
554 unsigned long ra_range_pages;
555 unsigned long ra_max_read_ahead_whole_pages;
556 struct workqueue_struct *ll_readahead_wq;
558 * Max number of active works could be triggered
559 * for async readahead.
561 unsigned int ra_async_max_active;
562 /* how many async readahead triggered in flight */
563 atomic_t ra_async_inflight;
564 /* Threshold to control when to trigger async readahead */
565 unsigned long ra_async_pages_per_file_threshold;
568 /* ra_io_arg will be filled in the beginning of ll_readahead with
569 * ras_lock, then the following ll_read_ahead_pages will read RA
570 * pages according to this arg, all the items in this structure are
571 * counted by page index.
574 pgoff_t ria_start_idx; /* start offset of read-ahead*/
575 pgoff_t ria_end_idx; /* end offset of read-ahead*/
576 unsigned long ria_reserved; /* reserved pages for read-ahead */
577 pgoff_t ria_end_idx_min;/* minimum end to cover current read */
578 bool ria_eof; /* reach end of file */
579 /* If stride read pattern is detected, ria_stoff is the byte offset
580 * where stride read is started. Note: for normal read-ahead, the
581 * value here is meaningless, and also it will not be accessed*/
583 /* ria_length and ria_bytes are the length and pages length in the
584 * stride I/O mode. And they will also be used to check whether
585 * it is stride I/O read-ahead in the read-ahead pages*/
590 /* LL_HIST_MAX=32 causes an overflow */
591 #define LL_HIST_MAX 28
592 #define LL_HIST_START 12 /* buckets start at 2^12 = 4k */
593 #define LL_PROCESS_HIST_MAX 10
594 struct per_process_info {
596 struct obd_histogram pp_r_hist;
597 struct obd_histogram pp_w_hist;
600 /* pp_extents[LL_PROCESS_HIST_MAX] will hold the combined process info */
601 struct ll_rw_extents_info {
602 struct per_process_info pp_extents[LL_PROCESS_HIST_MAX + 1];
605 #define LL_OFFSET_HIST_MAX 100
606 struct ll_rw_process_info {
609 loff_t rw_range_start;
611 loff_t rw_last_file_pos;
613 size_t rw_smallest_extent;
614 size_t rw_largest_extent;
615 struct ll_file_data *rw_last_file;
618 enum stats_track_type {
619 STATS_TRACK_ALL = 0, /* track all processes */
620 STATS_TRACK_PID, /* track process with this pid */
621 STATS_TRACK_PPID, /* track processes with this ppid */
622 STATS_TRACK_GID, /* track processes with this gid */
626 /* flags for sbi->ll_flags */
627 #define LL_SBI_NOLCK 0x01 /* DLM locking disabled (directio-only) */
628 #define LL_SBI_CHECKSUM 0x02 /* checksum each page as it's written */
629 #define LL_SBI_FLOCK 0x04
630 #define LL_SBI_USER_XATTR 0x08 /* support user xattr */
631 #define LL_SBI_ACL 0x10 /* support ACL */
632 /* LL_SBI_RMT_CLIENT 0x40 remote client */
633 #define LL_SBI_MDS_CAPA 0x80 /* support mds capa, obsolete */
634 #define LL_SBI_OSS_CAPA 0x100 /* support oss capa, obsolete */
635 #define LL_SBI_LOCALFLOCK 0x200 /* Local flocks support by kernel */
636 #define LL_SBI_LRU_RESIZE 0x400 /* lru resize support */
637 #define LL_SBI_LAZYSTATFS 0x800 /* lazystatfs mount option */
638 /* LL_SBI_SOM_PREVIEW 0x1000 SOM preview mount option, obsolete */
639 #define LL_SBI_32BIT_API 0x2000 /* generate 32 bit inodes. */
640 #define LL_SBI_64BIT_HASH 0x4000 /* support 64-bits dir hash/offset */
641 #define LL_SBI_AGL_ENABLED 0x8000 /* enable agl */
642 #define LL_SBI_VERBOSE 0x10000 /* verbose mount/umount */
643 #define LL_SBI_LAYOUT_LOCK 0x20000 /* layout lock support */
644 #define LL_SBI_USER_FID2PATH 0x40000 /* allow fid2path by unprivileged users */
645 #define LL_SBI_XATTR_CACHE 0x80000 /* support for xattr cache */
646 #define LL_SBI_NOROOTSQUASH 0x100000 /* do not apply root squash */
647 #define LL_SBI_ALWAYS_PING 0x200000 /* always ping even if server
649 #define LL_SBI_FAST_READ 0x400000 /* fast read support */
650 #define LL_SBI_FILE_SECCTX 0x800000 /* set file security context at create */
651 /* LL_SBI_PIO 0x1000000 parallel IO support, introduced in
653 #define LL_SBI_TINY_WRITE 0x2000000 /* tiny write support */
654 #define LL_SBI_FILE_HEAT 0x4000000 /* file heat support */
655 #define LL_SBI_TEST_DUMMY_ENCRYPTION 0x8000000 /* test dummy encryption */
656 #define LL_SBI_ENCRYPT 0x10000000 /* client side encryption */
657 #define LL_SBI_FOREIGN_SYMLINK 0x20000000 /* foreign fake-symlink support */
658 /* foreign fake-symlink upcall registered */
659 #define LL_SBI_FOREIGN_SYMLINK_UPCALL 0x40000000
660 #define LL_SBI_FLAGS { \
688 "test_dummy_encryption", \
691 "foreign_symlink_upcall", \
694 /* This is embedded into llite super-blocks to keep track of connect
695 * flags (capabilities) supported by all imports given mount is
697 struct lustre_client_ocd {
698 /* This is conjunction of connect_flags across all imports
699 * (LOVs) this mount is connected to. This field is updated by
700 * cl_ocd_update() under ->lco_lock. */
702 struct mutex lco_lock;
703 struct obd_export *lco_md_exp;
704 struct obd_export *lco_dt_exp;
708 /* this protects pglist and ra_info. It isn't safe to
709 * grab from interrupt contexts */
711 spinlock_t ll_pp_extent_lock; /* pp_extent entry*/
712 spinlock_t ll_process_lock; /* ll_rw_process_info */
713 struct obd_uuid ll_sb_uuid;
714 struct obd_export *ll_md_exp;
715 struct obd_export *ll_dt_exp;
716 struct obd_device *ll_md_obd;
717 struct obd_device *ll_dt_obd;
718 struct dentry *ll_debugfs_entry;
719 struct lu_fid ll_root_fid; /* root object fid */
722 unsigned int ll_xattr_cache_enabled:1,
723 ll_xattr_cache_set:1, /* already set to 0/1 */
724 ll_client_common_fill_super_succeeded:1,
727 struct lustre_client_ocd ll_lco;
729 struct lprocfs_stats *ll_stats; /* lprocfs stats counter */
731 /* Used to track "unstable" pages on a client, and maintain a
732 * LRU list of clean pages. An "unstable" page is defined as
733 * any page which is sent to a server as part of a bulk request,
734 * but is uncommitted to stable storage. */
735 struct cl_client_cache *ll_cache;
737 struct lprocfs_stats *ll_ra_stats;
739 struct ll_ra_info ll_ra_info;
740 unsigned int ll_namelen;
741 const struct file_operations *ll_fop;
743 struct lu_site *ll_site;
744 struct cl_device *ll_cl;
746 struct ll_rw_extents_info ll_rw_extents_info;
747 int ll_extent_process_count;
748 struct ll_rw_process_info ll_rw_process_info[LL_PROCESS_HIST_MAX];
749 unsigned int ll_offset_process_count;
750 struct ll_rw_process_info ll_rw_offset_info[LL_OFFSET_HIST_MAX];
751 unsigned int ll_rw_offset_entry_count;
752 int ll_stats_track_id;
753 enum stats_track_type ll_stats_track_type;
756 /* metadata stat-ahead */
757 unsigned int ll_sa_running_max;/* max concurrent
758 * statahead instances */
759 unsigned int ll_sa_max; /* max statahead RPCs */
760 atomic_t ll_sa_total; /* statahead thread started
762 atomic_t ll_sa_wrong; /* statahead thread stopped for
764 atomic_t ll_sa_running; /* running statahead thread
766 atomic_t ll_agl_total; /* AGL thread started count */
768 dev_t ll_sdev_orig; /* save s_dev before assign for
771 struct root_squash_info ll_squash;
774 /* st_blksize returned by stat(2), when non-zero */
775 unsigned int ll_stat_blksize;
777 /* maximum relative age of cached statfs results */
778 unsigned int ll_statfs_max_age;
780 struct kset ll_kset; /* sysfs object */
781 struct completion ll_kobj_unregister;
784 unsigned int ll_heat_decay_weight;
785 unsigned int ll_heat_period_second;
787 /* filesystem fsname */
788 char ll_fsname[LUSTRE_MAXFSNAME + 1];
790 /* Persistent Client Cache */
791 struct pcc_super ll_pcc_super;
793 /* to protect vs updates in all following foreign symlink fields */
794 struct rw_semaphore ll_foreign_symlink_sem;
795 /* foreign symlink path prefix */
796 char *ll_foreign_symlink_prefix;
797 /* full prefix size including leading '\0' */
798 size_t ll_foreign_symlink_prefix_size;
799 /* foreign symlink path upcall */
800 char *ll_foreign_symlink_upcall;
801 /* foreign symlink path upcall infos */
802 struct ll_foreign_symlink_upcall_item *ll_foreign_symlink_upcall_items;
803 /* foreign symlink path upcall nb infos */
804 unsigned int ll_foreign_symlink_upcall_nb_items;
807 #define SBI_DEFAULT_HEAT_DECAY_WEIGHT ((80 * 256 + 50) / 100)
808 #define SBI_DEFAULT_HEAT_PERIOD_SECOND (60)
810 * per file-descriptor read-ahead data.
812 struct ll_readahead_state {
814 /* End byte that read(2) try to read. */
815 loff_t ras_last_read_end_bytes;
817 * number of bytes read after last read-ahead window reset. As window
818 * is reset on each seek, this is effectively a number of consecutive
819 * accesses. Maybe ->ras_accessed_in_window is better name.
821 * XXX nikita: window is also reset (by ras_update()) when Lustre
822 * believes that memory pressure evicts read-ahead pages. In that
823 * case, it probably doesn't make sense to expand window to
824 * PTLRPC_MAX_BRW_PAGES on the third access.
826 loff_t ras_consecutive_bytes;
828 * number of read requests after the last read-ahead window reset
829 * As window is reset on each seek, this is effectively the number
830 * on consecutive read request and is used to trigger read-ahead.
832 unsigned long ras_consecutive_requests;
834 * Parameters of current read-ahead window. Handled by
835 * ras_update(). On the initial access to the file or after a seek,
836 * window is reset to 0. After 3 consecutive accesses, window is
837 * expanded to PTLRPC_MAX_BRW_PAGES. Afterwards, window is enlarged by
838 * PTLRPC_MAX_BRW_PAGES chunks up to ->ra_max_pages.
840 pgoff_t ras_window_start_idx;
841 pgoff_t ras_window_pages;
843 /* Page index where min range read starts */
844 pgoff_t ras_range_min_start_idx;
845 /* Page index where mmap range read ends */
846 pgoff_t ras_range_max_end_idx;
847 /* number of mmap pages where last time detected */
848 pgoff_t ras_last_range_pages;
849 /* number of mmap range requests */
850 pgoff_t ras_range_requests;
853 * Optimal RPC size in pages.
854 * It decides how many pages will be sent for each read-ahead.
856 unsigned long ras_rpc_pages;
858 * Where next read-ahead should start at. This lies within read-ahead
859 * window. Read-ahead window is read in pieces rather than at once
860 * because: 1. lustre limits total number of pages under read-ahead by
861 * ->ra_max_pages (see ll_ra_count_get()), 2. client cannot read pages
862 * not covered by DLM lock.
864 pgoff_t ras_next_readahead_idx;
866 * Total number of ll_file_read requests issued, reads originating
867 * due to mmap are not counted in this total. This value is used to
868 * trigger full file read-ahead after multiple reads to a small file.
870 unsigned long ras_requests;
872 * The following 3 items are used for detecting the stride I/O
874 * In stride I/O mode,
875 * ...............|-----data-----|****gap*****|--------|******|....
876 * offset |-stride_bytes-|-stride_gap-|
877 * ras_stride_offset = offset;
878 * ras_stride_length = stride_bytes + stride_gap;
879 * ras_stride_bytes = stride_bytes;
880 * Note: all these three items are counted by bytes.
882 loff_t ras_stride_offset;
883 loff_t ras_stride_length;
884 loff_t ras_stride_bytes;
886 * number of consecutive stride request count, and it is similar as
887 * ras_consecutive_requests, but used for stride I/O mode.
888 * Note: only more than 2 consecutive stride request are detected,
889 * stride read-ahead will be enable
891 unsigned long ras_consecutive_stride_requests;
892 /* index of the last page that async readahead starts */
893 pgoff_t ras_async_last_readpage_idx;
894 /* whether we should increase readahead window */
895 bool ras_need_increase_window;
896 /* whether ra miss check should be skipped */
897 bool ras_no_miss_check;
900 struct ll_readahead_work {
901 /** File to readahead */
902 struct file *lrw_file;
903 pgoff_t lrw_start_idx;
906 /* async worker to handler read */
907 struct work_struct lrw_readahead_work;
908 char lrw_jobid[LUSTRE_JOBID_SIZE];
911 extern struct kmem_cache *ll_file_data_slab;
912 struct lustre_handle;
913 struct ll_file_data {
914 struct ll_readahead_state fd_ras;
915 struct ll_grouplock fd_grouplock;
919 /* openhandle if lease exists for this file.
920 * Borrow lli->lli_och_mutex to protect assignment */
921 struct obd_client_handle *fd_lease_och;
922 struct obd_client_handle *fd_och;
923 struct file *fd_file;
924 /* Indicate whether need to report failure when close.
925 * true: failure is known, not report again.
926 * false: unknown failure, should report. */
927 bool fd_write_failed;
928 bool ll_lock_no_expand;
929 rwlock_t fd_lock; /* protect lcc list */
930 struct list_head fd_lccs; /* list of ll_cl_context */
931 /* Used by mirrored file to lead IOs to a specific mirror, usually
932 * for mirror resync. 0 means default. */
933 __u32 fd_designated_mirror;
934 /* The layout version when resync starts. Resync I/O should carry this
935 * layout version for verification to OST objects */
936 __u32 fd_layout_version;
937 struct pcc_file fd_pcc_file;
940 void llite_tunables_unregister(void);
941 int llite_tunables_register(void);
943 static inline struct inode *ll_info2i(struct ll_inode_info *lli)
945 return &lli->lli_vfs_inode;
948 __u32 ll_i2suppgid(struct inode *i);
949 void ll_i2gids(__u32 *suppgids, struct inode *i1,struct inode *i2);
951 static inline int ll_need_32bit_api(struct ll_sb_info *sbi)
953 #if BITS_PER_LONG == 32
955 #elif defined(CONFIG_COMPAT)
956 if (unlikely(sbi->ll_flags & LL_SBI_32BIT_API))
959 # ifdef CONFIG_X86_X32
960 /* in_compat_syscall() returns true when called from a kthread
961 * and CONFIG_X86_X32 is enabled, which is wrong. So check
962 * whether the caller comes from a syscall (ie. not a kthread)
963 * before calling in_compat_syscall(). */
964 if (current->flags & PF_KTHREAD)
968 return unlikely(in_compat_syscall());
970 return unlikely(sbi->ll_flags & LL_SBI_32BIT_API);
974 static inline bool ll_sbi_has_fast_read(struct ll_sb_info *sbi)
976 return !!(sbi->ll_flags & LL_SBI_FAST_READ);
979 static inline bool ll_sbi_has_tiny_write(struct ll_sb_info *sbi)
981 return !!(sbi->ll_flags & LL_SBI_TINY_WRITE);
984 static inline bool ll_sbi_has_file_heat(struct ll_sb_info *sbi)
986 return !!(sbi->ll_flags & LL_SBI_FILE_HEAT);
989 static inline bool ll_sbi_has_foreign_symlink(struct ll_sb_info *sbi)
991 return !!(sbi->ll_flags & LL_SBI_FOREIGN_SYMLINK);
994 void ll_ras_enter(struct file *f, loff_t pos, size_t count);
996 /* llite/lcommon_misc.c */
997 int cl_ocd_update(struct obd_device *host, struct obd_device *watched,
998 enum obd_notify_event ev, void *owner);
999 int cl_get_grouplock(struct cl_object *obj, unsigned long gid, int nonblock,
1000 struct ll_grouplock *lg);
1001 void cl_put_grouplock(struct ll_grouplock *lg);
1003 /* llite/lproc_llite.c */
1004 int ll_debugfs_register_super(struct super_block *sb, const char *name);
1005 void ll_debugfs_unregister_super(struct super_block *sb);
1006 void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, long count);
1009 LPROC_LL_READ_BYTES,
1010 LPROC_LL_WRITE_BYTES,
1037 LPROC_LL_GETXATTR_HITS,
1039 LPROC_LL_REMOVEXATTR,
1040 LPROC_LL_INODE_PERM,
1042 LPROC_LL_FILE_OPCODES
1046 enum get_default_layout_type {
1047 GET_DEFAULT_LAYOUT_ROOT = 1,
1050 extern const struct file_operations ll_dir_operations;
1051 extern const struct inode_operations ll_dir_inode_operations;
1052 #ifdef HAVE_DIR_CONTEXT
1053 int ll_dir_read(struct inode *inode, __u64 *pos, struct md_op_data *op_data,
1054 struct dir_context *ctx);
1056 int ll_dir_read(struct inode *inode, __u64 *pos, struct md_op_data *op_data,
1057 void *cookie, filldir_t filldir);
1059 int ll_get_mdt_idx(struct inode *inode);
1060 int ll_get_mdt_idx_by_fid(struct ll_sb_info *sbi, const struct lu_fid *fid);
1061 struct page *ll_get_dir_page(struct inode *dir, struct md_op_data *op_data,
1063 void ll_release_page(struct inode *inode, struct page *page, bool remove);
1064 int quotactl_ioctl(struct ll_sb_info *sbi, struct if_quotactl *qctl);
1067 extern const struct inode_operations ll_special_inode_operations;
1069 struct inode *ll_iget(struct super_block *sb, ino_t hash,
1070 struct lustre_md *lic);
1071 int ll_test_inode_by_fid(struct inode *inode, void *opaque);
1072 int ll_md_blocking_ast(struct ldlm_lock *, struct ldlm_lock_desc *,
1073 void *data, int flag);
1074 struct dentry *ll_splice_alias(struct inode *inode, struct dentry *de);
1075 int ll_rmdir_entry(struct inode *dir, char *name, int namelen);
1076 void ll_update_times(struct ptlrpc_request *request, struct inode *inode);
1079 int ll_writepage(struct page *page, struct writeback_control *wbc);
1080 int ll_writepages(struct address_space *, struct writeback_control *wbc);
1081 int ll_readpage(struct file *file, struct page *page);
1082 int ll_io_read_page(const struct lu_env *env, struct cl_io *io,
1083 struct cl_page *page, struct file *file);
1084 void ll_readahead_init(struct inode *inode, struct ll_readahead_state *ras);
1085 int vvp_io_write_commit(const struct lu_env *env, struct cl_io *io);
1088 void ll_cl_add(struct file *file, const struct lu_env *env, struct cl_io *io,
1089 enum lcc_type type);
1090 void ll_cl_remove(struct file *file, const struct lu_env *env);
1091 struct ll_cl_context *ll_cl_find(struct file *file);
1093 extern const struct address_space_operations ll_aops;
1096 extern const struct file_operations ll_file_operations;
1097 extern const struct file_operations ll_file_operations_flock;
1098 extern const struct file_operations ll_file_operations_noflock;
1099 extern const struct inode_operations ll_file_inode_operations;
1100 extern int ll_have_md_lock(struct inode *inode, __u64 *bits,
1101 enum ldlm_mode l_req_mode);
1102 extern enum ldlm_mode ll_take_md_lock(struct inode *inode, __u64 bits,
1103 struct lustre_handle *lockh, __u64 flags,
1104 enum ldlm_mode mode);
1106 int ll_file_open(struct inode *inode, struct file *file);
1107 int ll_file_release(struct inode *inode, struct file *file);
1108 int ll_release_openhandle(struct dentry *, struct lookup_intent *);
1109 int ll_md_real_close(struct inode *inode, fmode_t fmode);
1110 extern void ll_rw_stats_tally(struct ll_sb_info *sbi, pid_t pid,
1111 struct ll_file_data *file, loff_t pos,
1112 size_t count, int rw);
1113 #ifdef HAVE_INODEOPS_ENHANCED_GETATTR
1114 int ll_getattr(const struct path *path, struct kstat *stat,
1115 u32 request_mask, unsigned int flags);
1117 int ll_getattr(struct vfsmount *mnt, struct dentry *de, struct kstat *stat);
1119 int ll_getattr_dentry(struct dentry *de, struct kstat *stat, u32 request_mask,
1120 unsigned int flags, bool foreign);
1121 #ifdef CONFIG_LUSTRE_FS_POSIX_ACL
1122 struct posix_acl *ll_get_acl(struct inode *inode, int type);
1123 int ll_set_acl(struct inode *inode, struct posix_acl *acl, int type);
1124 #else /* !CONFIG_LUSTRE_FS_POSIX_ACL */
1125 #define ll_get_acl NULL
1126 #define ll_set_acl NULL
1127 #endif /* CONFIG_LUSTRE_FS_POSIX_ACL */
1129 static inline int ll_xflags_to_inode_flags(int xflags)
1131 return ((xflags & FS_XFLAG_SYNC) ? S_SYNC : 0) |
1132 ((xflags & FS_XFLAG_NOATIME) ? S_NOATIME : 0) |
1133 ((xflags & FS_XFLAG_APPEND) ? S_APPEND : 0) |
1134 ((xflags & FS_XFLAG_IMMUTABLE) ? S_IMMUTABLE : 0);
1137 static inline int ll_inode_flags_to_xflags(int inode_flags)
1139 return ((inode_flags & S_SYNC) ? FS_XFLAG_SYNC : 0) |
1140 ((inode_flags & S_NOATIME) ? FS_XFLAG_NOATIME : 0) |
1141 ((inode_flags & S_APPEND) ? FS_XFLAG_APPEND : 0) |
1142 ((inode_flags & S_IMMUTABLE) ? FS_XFLAG_IMMUTABLE : 0);
1145 int ll_migrate(struct inode *parent, struct file *file,
1146 struct lmv_user_md *lum, const char *name);
1147 int ll_get_fid_by_name(struct inode *parent, const char *name,
1148 int namelen, struct lu_fid *fid, struct inode **inode);
1149 int ll_inode_permission(struct inode *inode, int mask);
1150 int ll_ioctl_check_project(struct inode *inode, struct fsxattr *fa);
1151 int ll_ioctl_fsgetxattr(struct inode *inode, unsigned int cmd,
1153 int ll_ioctl_fssetxattr(struct inode *inode, unsigned int cmd,
1156 int ll_lov_setstripe_ea_info(struct inode *inode, struct dentry *dentry,
1157 __u64 flags, struct lov_user_md *lum,
1159 int ll_lov_getstripe_ea_info(struct inode *inode, const char *filename,
1160 struct lov_mds_md **lmm, int *lmm_size,
1161 struct ptlrpc_request **request);
1162 int ll_dir_setstripe(struct inode *inode, struct lov_user_md *lump,
1164 int ll_dir_getstripe_default(struct inode *inode, void **lmmp,
1165 int *lmm_size, struct ptlrpc_request **request,
1166 struct ptlrpc_request **root_request, u64 valid);
1167 int ll_dir_getstripe(struct inode *inode, void **plmm, int *plmm_size,
1168 struct ptlrpc_request **request, u64 valid);
1169 int ll_fsync(struct file *file, loff_t start, loff_t end, int data);
1170 int ll_merge_attr(const struct lu_env *env, struct inode *inode);
1171 int ll_fid2path(struct inode *inode, void __user *arg);
1172 int ll_data_version(struct inode *inode, __u64 *data_version, int flags);
1173 int ll_hsm_release(struct inode *inode);
1174 int ll_hsm_state_set(struct inode *inode, struct hsm_state_set *hss);
1175 void ll_io_set_mirror(struct cl_io *io, const struct file *file);
1177 /* llite/dcache.c */
1179 int ll_d_init(struct dentry *de);
1180 extern const struct dentry_operations ll_d_ops;
1181 void ll_intent_drop_lock(struct lookup_intent *);
1182 void ll_intent_release(struct lookup_intent *);
1183 void ll_prune_aliases(struct inode *inode);
1184 void ll_lookup_finish_locks(struct lookup_intent *it, struct dentry *dentry);
1185 int ll_revalidate_it_finish(struct ptlrpc_request *request,
1186 struct lookup_intent *it, struct dentry *de);
1188 /* llite/llite_lib.c */
1189 extern const struct super_operations lustre_super_operations;
1191 void ll_lli_init(struct ll_inode_info *lli);
1192 int ll_fill_super(struct super_block *sb);
1193 void ll_put_super(struct super_block *sb);
1194 void ll_kill_super(struct super_block *sb);
1195 struct inode *ll_inode_from_resource_lock(struct ldlm_lock *lock);
1196 void ll_dir_clear_lsm_md(struct inode *inode);
1197 void ll_clear_inode(struct inode *inode);
1198 int ll_setattr_raw(struct dentry *dentry, struct iattr *attr,
1199 enum op_xvalid xvalid, bool hsm_import);
1200 int ll_setattr(struct dentry *de, struct iattr *attr);
1201 int ll_statfs(struct dentry *de, struct kstatfs *sfs);
1202 int ll_statfs_internal(struct ll_sb_info *sbi, struct obd_statfs *osfs,
1204 int ll_update_inode(struct inode *inode, struct lustre_md *md);
1205 void ll_update_inode_flags(struct inode *inode, unsigned int ext_flags);
1206 int ll_read_inode2(struct inode *inode, void *opaque);
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 ptlrpc_request *req,
1216 struct super_block *, struct lookup_intent *);
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 ptlrpc_request *open_req);
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;
1297 struct ll_thread_info {
1298 struct vvp_io_args lti_args;
1299 struct ra_io_arg lti_ria;
1300 struct ll_cl_context lti_io_ctx;
1303 extern struct lu_context_key ll_thread_key;
1305 static inline struct ll_thread_info *ll_env_info(const struct lu_env *env)
1307 struct ll_thread_info *lti;
1309 lti = lu_context_key_get(&env->le_ctx, &ll_thread_key);
1310 LASSERT(lti != NULL);
1315 static inline struct vvp_io_args *ll_env_args(const struct lu_env *env)
1317 return &ll_env_info(env)->lti_args;
1320 void ll_io_init(struct cl_io *io, struct file *file, enum cl_io_type iot,
1321 struct vvp_io_args *args);
1323 /* llite/llite_mmap.c */
1325 int ll_file_mmap(struct file * file, struct vm_area_struct * vma);
1326 void policy_from_vma(union ldlm_policy_data *policy, struct vm_area_struct *vma,
1327 unsigned long addr, size_t count);
1328 struct vm_area_struct *our_vma(struct mm_struct *mm, unsigned long addr,
1331 #define ll_s2sbi(sb) (s2lsi(sb)->lsi_llsbi)
1333 /* don't need an addref as the sb_info should be holding one */
1334 static inline struct obd_export *ll_s2dtexp(struct super_block *sb)
1336 return ll_s2sbi(sb)->ll_dt_exp;
1339 /* don't need an addref as the sb_info should be holding one */
1340 static inline struct obd_export *ll_s2mdexp(struct super_block *sb)
1342 return ll_s2sbi(sb)->ll_md_exp;
1345 static inline struct client_obd *sbi2mdc(struct ll_sb_info *sbi)
1347 struct obd_device *obd = sbi->ll_md_exp->exp_obd;
1353 // FIXME: replace the name of this with LL_SB to conform to kernel stuff
1354 static inline struct ll_sb_info *ll_i2sbi(struct inode *inode)
1356 return ll_s2sbi(inode->i_sb);
1359 static inline struct obd_export *ll_i2dtexp(struct inode *inode)
1361 return ll_s2dtexp(inode->i_sb);
1364 static inline struct obd_export *ll_i2mdexp(struct inode *inode)
1366 return ll_s2mdexp(inode->i_sb);
1369 static inline struct lu_fid *ll_inode2fid(struct inode *inode)
1373 LASSERT(inode != NULL);
1374 fid = &ll_i2info(inode)->lli_fid;
1379 static inline bool ll_dir_striped(struct inode *inode)
1382 return S_ISDIR(inode->i_mode) &&
1383 lmv_dir_striped(ll_i2info(inode)->lli_lsm_md);
1386 static inline loff_t ll_file_maxbytes(struct inode *inode)
1388 struct cl_object *obj = ll_i2info(inode)->lli_clob;
1391 return MAX_LFS_FILESIZE;
1393 return min_t(loff_t, cl_object_maxbytes(obj), MAX_LFS_FILESIZE);
1397 extern const struct xattr_handler *ll_xattr_handlers[];
1399 #define XATTR_USER_T 1
1400 #define XATTR_TRUSTED_T 2
1401 #define XATTR_SECURITY_T 3
1402 #define XATTR_ACL_ACCESS_T 4
1403 #define XATTR_ACL_DEFAULT_T 5
1404 #define XATTR_LUSTRE_T 6
1405 #define XATTR_OTHER_T 7
1407 ssize_t ll_listxattr(struct dentry *dentry, char *buffer, size_t size);
1408 int ll_xattr_list(struct inode *inode, const char *name, int type,
1409 void *buffer, size_t size, u64 valid);
1410 const struct xattr_handler *get_xattr_type(const char *name);
1413 * Common IO arguments for various VFS I/O interfaces.
1415 int cl_sb_init(struct super_block *sb);
1416 int cl_sb_fini(struct super_block *sb);
1418 enum ras_update_flags {
1422 void ll_ra_count_put(struct ll_sb_info *sbi, unsigned long len);
1423 void ll_ra_stats_inc(struct inode *inode, enum ra_stat which);
1427 #define LL_SA_RPC_MIN 2
1428 #define LL_SA_RPC_DEF 32
1429 #define LL_SA_RPC_MAX 512
1431 /* XXX: If want to support more concurrent statahead instances,
1432 * please consider to decentralize the RPC lists attached
1433 * on related import, such as imp_{sending,delayed}_list.
1435 #define LL_SA_RUNNING_MAX 256
1436 #define LL_SA_RUNNING_DEF 16
1438 #define LL_SA_CACHE_BIT 5
1439 #define LL_SA_CACHE_SIZE (1 << LL_SA_CACHE_BIT)
1440 #define LL_SA_CACHE_MASK (LL_SA_CACHE_SIZE - 1)
1442 /* per inode struct, for dir only */
1443 struct ll_statahead_info {
1444 struct dentry *sai_dentry;
1445 atomic_t sai_refcount; /* when access this struct, hold
1447 unsigned int sai_max; /* max ahead of lookup */
1448 __u64 sai_sent; /* stat requests sent count */
1449 __u64 sai_replied; /* stat requests which received
1451 __u64 sai_index; /* index of statahead entry */
1452 __u64 sai_index_wait; /* index of entry which is the
1453 * caller is waiting for */
1454 __u64 sai_hit; /* hit count */
1455 __u64 sai_miss; /* miss count:
1456 * for "ls -al" case, includes
1457 * hidden dentry miss;
1458 * for "ls -l" case, it does not
1459 * include hidden dentry miss.
1460 * "sai_miss_hidden" is used for
1463 unsigned int sai_consecutive_miss; /* consecutive miss */
1464 unsigned int sai_miss_hidden;/* "ls -al", but first dentry
1465 * is not a hidden one */
1466 unsigned int sai_skip_hidden;/* skipped hidden dentry count
1468 unsigned int sai_ls_all:1, /* "ls -al", do stat-ahead for
1470 sai_in_readpage:1;/* statahead is in readdir()*/
1471 wait_queue_head_t sai_waitq; /* stat-ahead wait queue */
1472 struct task_struct *sai_task; /* stat-ahead thread */
1473 struct task_struct *sai_agl_task; /* AGL thread */
1474 struct list_head sai_interim_entries; /* entries which got async
1475 * stat reply, but not
1477 struct list_head sai_entries; /* completed entries */
1478 struct list_head sai_agls; /* AGLs to be sent */
1479 struct list_head sai_cache[LL_SA_CACHE_SIZE];
1480 spinlock_t sai_cache_lock[LL_SA_CACHE_SIZE];
1481 atomic_t sai_cache_count; /* entry count in cache */
1484 int ll_revalidate_statahead(struct inode *dir, struct dentry **dentry,
1486 int ll_start_statahead(struct inode *dir, struct dentry *dentry, bool agl);
1487 void ll_authorize_statahead(struct inode *dir, void *key);
1488 void ll_deauthorize_statahead(struct inode *dir, void *key);
1491 blkcnt_t dirty_cnt(struct inode *inode);
1493 int cl_glimpse_size0(struct inode *inode, int agl);
1494 int cl_glimpse_lock(const struct lu_env *env, struct cl_io *io,
1495 struct inode *inode, struct cl_object *clob, int agl);
1497 static inline int cl_glimpse_size(struct inode *inode)
1499 return cl_glimpse_size0(inode, 0);
1502 /* AGL is 'asychronous glimpse lock', which is a speculative lock taken as
1503 * part of statahead */
1504 static inline int cl_agl(struct inode *inode)
1506 return cl_glimpse_size0(inode, 1);
1509 int ll_file_lock_ahead(struct file *file, struct llapi_lu_ladvise *ladvise);
1511 int cl_io_get(struct inode *inode, struct lu_env **envout,
1512 struct cl_io **ioout, __u16 *refcheck);
1514 static inline int ll_glimpse_size(struct inode *inode)
1516 struct ll_inode_info *lli = ll_i2info(inode);
1519 down_read(&lli->lli_glimpse_sem);
1520 rc = cl_glimpse_size(inode);
1521 lli->lli_glimpse_time = ktime_get();
1522 up_read(&lli->lli_glimpse_sem);
1526 /* dentry may statahead when statahead is enabled and current process has opened
1527 * parent directory, and this dentry hasn't accessed statahead cache before */
1529 dentry_may_statahead(struct inode *dir, struct dentry *dentry)
1531 struct ll_inode_info *lli;
1532 struct ll_dentry_data *ldd;
1534 if (ll_i2sbi(dir)->ll_sa_max == 0)
1537 lli = ll_i2info(dir);
1539 /* statahead is not allowed for this dir, there may be three causes:
1540 * 1. dir is not opened.
1541 * 2. statahead hit ratio is too low.
1542 * 3. previous stat started statahead thread failed. */
1543 if (!lli->lli_sa_enabled)
1546 /* not the same process, don't statahead */
1547 if (lli->lli_opendir_pid != current->pid)
1551 * When stating a dentry, kernel may trigger 'revalidate' or 'lookup'
1552 * multiple times, eg. for 'getattr', 'getxattr' and etc.
1553 * For patchless client, lookup intent is not accurate, which may
1554 * misguide statahead. For example:
1555 * The 'revalidate' call for 'getattr' and 'getxattr' of a dentry will
1556 * have the same intent -- IT_GETATTR, while one dentry should access
1557 * statahead cache once, otherwise statahead windows is messed up.
1558 * The solution is as following:
1559 * Assign 'lld_sa_generation' with 'lli_sa_generation' when a dentry
1560 * IT_GETATTR for the first time, and subsequent IT_GETATTR will
1561 * bypass interacting with statahead cache by checking
1562 * 'lld_sa_generation == lli->lli_sa_generation'.
1564 ldd = ll_d2d(dentry);
1565 if (ldd != NULL && lli->lli_sa_generation &&
1566 ldd->lld_sa_generation == lli->lli_sa_generation)
1572 int cl_sync_file_range(struct inode *inode, loff_t start, loff_t end,
1573 enum cl_fsync_mode mode, int ignore_layout);
1575 static inline int ll_file_nolock(const struct file *file)
1577 struct ll_file_data *fd = file->private_data;
1578 struct inode *inode = file_inode((struct file *)file);
1580 LASSERT(fd != NULL);
1581 return ((fd->fd_flags & LL_FILE_IGNORE_LOCK) ||
1582 (ll_i2sbi(inode)->ll_flags & LL_SBI_NOLCK));
1585 static inline void ll_set_lock_data(struct obd_export *exp, struct inode *inode,
1586 struct lookup_intent *it, __u64 *bits)
1588 if (!it->it_lock_set) {
1589 struct lustre_handle handle;
1591 /* If this inode is a remote object, it will get two
1592 * separate locks in different namespaces, Master MDT,
1593 * where the name entry is, will grant LOOKUP lock,
1594 * remote MDT, where the object is, will grant
1595 * UPDATE|PERM lock. The inode will be attched to both
1596 * LOOKUP and PERM locks, so revoking either locks will
1597 * case the dcache being cleared */
1598 if (it->it_remote_lock_mode) {
1599 handle.cookie = it->it_remote_lock_handle;
1600 CDEBUG(D_DLMTRACE, "setting l_data to inode "DFID
1601 "(%p) for remote lock %#llx\n",
1602 PFID(ll_inode2fid(inode)), inode,
1604 md_set_lock_data(exp, &handle, inode, NULL);
1607 handle.cookie = it->it_lock_handle;
1609 CDEBUG(D_DLMTRACE, "setting l_data to inode "DFID"(%p)"
1610 " for lock %#llx\n",
1611 PFID(ll_inode2fid(inode)), inode, handle.cookie);
1613 md_set_lock_data(exp, &handle, inode, &it->it_lock_bits);
1614 it->it_lock_set = 1;
1618 *bits = it->it_lock_bits;
1621 static inline int d_lustre_invalid(const struct dentry *dentry)
1623 struct ll_dentry_data *lld = ll_d2d(dentry);
1625 return (lld == NULL) || lld->lld_invalid;
1628 static inline void __d_lustre_invalidate(struct dentry *dentry)
1630 struct ll_dentry_data *lld = ll_d2d(dentry);
1633 lld->lld_invalid = 1;
1637 * Mark dentry INVALID, if dentry refcount is zero (this is normally case for
1638 * ll_md_blocking_ast), it will be pruned by ll_prune_aliases() and
1639 * ll_prune_negative_children(); otherwise dput() of the last refcount will
1640 * unhash this dentry and kill it.
1642 static inline void d_lustre_invalidate(struct dentry *dentry)
1644 CDEBUG(D_DENTRY, "invalidate dentry %pd (%p) parent %p inode %p refc %d\n",
1646 dentry->d_parent, dentry->d_inode, ll_d_count(dentry));
1648 spin_lock(&dentry->d_lock);
1649 __d_lustre_invalidate(dentry);
1650 spin_unlock(&dentry->d_lock);
1653 static inline void d_lustre_revalidate(struct dentry *dentry)
1655 spin_lock(&dentry->d_lock);
1656 LASSERT(ll_d2d(dentry) != NULL);
1657 ll_d2d(dentry)->lld_invalid = 0;
1658 spin_unlock(&dentry->d_lock);
1661 static inline dev_t ll_compat_encode_dev(dev_t dev)
1663 /* The compat_sys_*stat*() syscalls will fail unless the
1664 * device majors and minors are both less than 256. Note that
1665 * the value returned here will be passed through
1666 * old_encode_dev() in cp_compat_stat(). And so we are not
1667 * trying to return a valid compat (u16) device number, just
1668 * one that will pass the old_valid_dev() check. */
1670 return MKDEV(MAJOR(dev) & 0xff, MINOR(dev) & 0xff);
1673 int ll_layout_conf(struct inode *inode, const struct cl_object_conf *conf);
1674 int ll_layout_refresh(struct inode *inode, __u32 *gen);
1675 int ll_layout_restore(struct inode *inode, loff_t start, __u64 length);
1676 int ll_layout_write_intent(struct inode *inode, enum layout_intent_opc opc,
1677 struct lu_extent *ext);
1679 int ll_xattr_init(void);
1680 void ll_xattr_fini(void);
1682 int ll_page_sync_io(const struct lu_env *env, struct cl_io *io,
1683 struct cl_page *page, enum cl_req_type crt);
1685 int ll_getparent(struct file *file, struct getparent __user *arg);
1688 int cl_setattr_ost(struct cl_object *obj, const struct iattr *attr,
1689 enum op_xvalid xvalid, unsigned int attr_flags);
1691 extern struct lu_env *cl_inode_fini_env;
1692 extern __u16 cl_inode_fini_refcheck;
1694 int cl_file_inode_init(struct inode *inode, struct lustre_md *md);
1695 void cl_inode_fini(struct inode *inode);
1697 u64 cl_fid_build_ino(const struct lu_fid *fid, int api32);
1698 u32 cl_fid_build_gen(const struct lu_fid *fid);
1700 static inline struct pcc_super *ll_i2pccs(struct inode *inode)
1702 return &ll_i2sbi(inode)->ll_pcc_super;
1705 static inline struct pcc_super *ll_info2pccs(struct ll_inode_info *lli)
1707 return ll_i2pccs(ll_info2i(lli));
1710 #ifdef HAVE_LUSTRE_CRYPTO
1712 extern const struct llcrypt_operations lustre_cryptops;
1714 /* llite/llite_foreign.c */
1715 int ll_manage_foreign(struct inode *inode, struct lustre_md *lmd);
1716 bool ll_foreign_is_openable(struct dentry *dentry, unsigned int flags);
1717 bool ll_foreign_is_removable(struct dentry *dentry, bool unset);
1719 #endif /* LLITE_INTERNAL_H */