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, 2015, 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
35 #include <lustre_debug.h>
36 #include <uapi/linux/lustre/lustre_ver.h>
37 #include <lustre_disk.h> /* for s2sbi */
38 #include <lustre_linkea.h>
40 /* for struct cl_lock_descr and struct cl_io */
41 #include <lustre_patchless_compat.h>
42 #include <lustre_compat.h>
43 #include <cl_object.h>
44 #include <lustre_lmv.h>
45 #include <lustre_mdc.h>
46 #include <lustre_intent.h>
47 #include <linux/compat.h>
48 #include <linux/namei.h>
49 #include <linux/xattr.h>
50 #include <linux/posix_acl_xattr.h>
51 #include "vvp_internal.h"
52 #include "range_lock.h"
58 #ifndef VM_FAULT_RETRY
59 #define VM_FAULT_RETRY 0
62 /** Only used on client-side for indicating the tail of dir hash/offset. */
63 #define LL_DIR_END_OFF 0x7fffffffffffffffULL
64 #define LL_DIR_END_OFF_32BIT 0x7fffffffUL
66 /* 4UL * 1024 * 1024 */
67 #define LL_MAX_BLKSIZE_BITS 22
69 #define LL_IT2STR(it) ((it) ? ldlm_it2str((it)->it_op) : "0")
70 #define LUSTRE_FPRIVATE(file) ((file)->private_data)
72 struct ll_dentry_data {
73 struct lookup_intent *lld_it;
74 unsigned int lld_sa_generation;
75 unsigned int lld_invalid:1;
76 unsigned int lld_nfs_dentry:1;
77 struct rcu_head lld_rcu_head;
80 #define ll_d2d(de) ((struct ll_dentry_data *)((de)->d_fsdata))
82 #define LLI_INODE_MAGIC 0x111d0de5
83 #define LLI_INODE_DEAD 0xdeadd00d
85 struct ll_getname_data {
86 struct dir_context ctx;
87 char *lgd_name; /* points to a buffer with NAME_MAX+1 size */
88 struct lu_fid lgd_fid; /* target fid we are looking for */
89 int lgd_found; /* inode matched? */
93 struct lu_env *lg_env;
95 struct cl_lock *lg_lock;
100 /* File data is modified. */
101 LLIF_DATA_MODIFIED = 0,
102 /* File is being restored */
103 LLIF_FILE_RESTORING = 1,
104 /* Xattr cache is attached to the file */
105 LLIF_XATTR_CACHE = 2,
108 struct ll_inode_info {
109 __u32 lli_inode_magic;
112 unsigned long lli_flags;
113 struct posix_acl *lli_posix_acl;
115 /* identifying fields for both metadata and data stacks. */
116 struct lu_fid lli_fid;
117 /* master inode fid for stripe directory */
118 struct lu_fid lli_pfid;
120 /* We need all three because every inode may be opened in different
123 struct obd_client_handle *lli_mds_read_och;
124 struct obd_client_handle *lli_mds_write_och;
125 struct obd_client_handle *lli_mds_exec_och;
126 __u64 lli_open_fd_read_count;
127 __u64 lli_open_fd_write_count;
128 __u64 lli_open_fd_exec_count;
129 /* Protects access to och pointers and their usage counters */
130 struct mutex lli_och_mutex;
132 struct inode lli_vfs_inode;
134 /* the most recent timestamps obtained from mds */
138 spinlock_t lli_agl_lock;
140 /* Try to make the d::member and f::member are aligned. Before using
141 * these members, make clear whether it is directory or not.
146 /* serialize normal readdir and statahead-readdir. */
147 struct mutex lli_readdir_mutex;
149 /* metadata statahead */
150 /* since parent-child threads can share the same @file
151 * struct, "opendir_key" is the token when dir close for
152 * case of parent exit before child -- it is me should
153 * cleanup the dir readahead.
155 void *lli_opendir_key;
156 struct ll_statahead_info *lli_sai;
157 /* protect statahead stuff. */
158 spinlock_t lli_sa_lock;
159 /* "opendir_pid" is the token when lookup/revalidate
160 * -- I am the owner of dir statahead.
162 pid_t lli_opendir_pid;
163 /* stat will try to access statahead entries or start
164 * statahead if this flag is set, and this flag will be
165 * set upon dir open, and cleared when dir is closed,
166 * statahead hit ratio is too low, or start statahead
169 unsigned int lli_sa_enabled:1;
170 /* generation for statahead */
171 unsigned int lli_sa_generation;
172 /* directory stripe information */
173 struct lmv_stripe_md *lli_lsm_md;
174 /* default directory stripe offset. This is extracted
175 * from the "dmv" xattr in order to decide which MDT to
176 * create a subdirectory on. The MDS itself fetches
177 * "dmv" and gets the rest of the default layout itself
178 * (count, hash, etc).
180 __u32 lli_def_stripe_offset;
183 /* for non-directory */
185 struct mutex lli_size_mutex;
186 char *lli_symlink_name;
188 * struct rw_semaphore {
189 * signed long count; // align d.d_def_acl
190 * spinlock_t wait_lock; // align d.d_sa_lock
191 * struct list_head wait_list;
194 struct rw_semaphore lli_trunc_sem;
195 struct range_lock_tree lli_write_tree;
197 struct rw_semaphore lli_glimpse_sem;
198 unsigned long lli_glimpse_time;
199 struct list_head lli_agl_list;
202 /* for writepage() only to communicate to fsync */
206 * whenever a process try to read/write the file, the
207 * jobid of the process will be saved here, and it'll
208 * be packed into the write PRC when flush later.
210 * so the read/write statistics for jobid will not be
211 * accurate if the file is shared by different jobs.
213 char lli_jobid[LUSTRE_JOBID_SIZE];
217 /* XXX: For following frequent used members, although they maybe special
218 * used for non-directory object, it is some time-wasting to check
219 * whether the object is directory or not before using them. On the
220 * other hand, currently, sizeof(f) > sizeof(d), it cannot reduce
221 * the "ll_inode_info" size even if moving those members into u.f.
222 * So keep them out side.
224 * In the future, if more members are added only for directory,
225 * some of the following members can be moved into u.f.
227 struct cl_object *lli_clob;
229 /* mutex to request for layout lock exclusively. */
230 struct mutex lli_layout_mutex;
231 /* Layout version, protected by lli_layout_lock */
232 __u32 lli_layout_gen;
233 spinlock_t lli_layout_lock;
235 struct rw_semaphore lli_xattrs_list_rwsem;
236 struct mutex lli_xattrs_enq_lock;
237 struct list_head lli_xattrs;/* ll_xattr_entry->xe_list */
240 static inline __u32 ll_layout_version_get(struct ll_inode_info *lli)
244 spin_lock(&lli->lli_layout_lock);
245 gen = lli->lli_layout_gen;
246 spin_unlock(&lli->lli_layout_lock);
251 static inline void ll_layout_version_set(struct ll_inode_info *lli, __u32 gen)
253 spin_lock(&lli->lli_layout_lock);
254 lli->lli_layout_gen = gen;
255 spin_unlock(&lli->lli_layout_lock);
258 int ll_xattr_cache_destroy(struct inode *inode);
260 int ll_xattr_cache_get(struct inode *inode, const char *name,
261 char *buffer, size_t size, __u64 valid);
263 int ll_init_security(struct dentry *dentry, struct inode *inode,
267 * Locking to guarantee consistency of non-atomic updates to long long i_size,
268 * consistency between file size and KMS.
270 * Implemented by ->lli_size_mutex and ->lsm_lock, nested in that order.
273 void ll_inode_size_lock(struct inode *inode);
274 void ll_inode_size_unlock(struct inode *inode);
276 /* FIXME: replace the name of this with LL_I to conform to kernel stuff */
277 /* static inline struct ll_inode_info *LL_I(struct inode *inode) */
278 static inline struct ll_inode_info *ll_i2info(struct inode *inode)
280 return container_of(inode, struct ll_inode_info, lli_vfs_inode);
283 /* default to about 64M of readahead on a given system. */
284 #define SBI_DEFAULT_READAHEAD_MAX (64UL << (20 - PAGE_SHIFT))
286 /* default to read-ahead full files smaller than 2MB on the second read */
287 #define SBI_DEFAULT_READAHEAD_WHOLE_MAX (2UL << (20 - PAGE_SHIFT))
292 RA_STAT_DISTANT_READPAGE,
293 RA_STAT_MISS_IN_WINDOW,
294 RA_STAT_FAILED_GRAB_PAGE,
295 RA_STAT_FAILED_MATCH,
300 RA_STAT_MAX_IN_FLIGHT,
301 RA_STAT_WRONG_GRAB_PAGE,
302 RA_STAT_FAILED_REACH_END,
307 atomic_t ra_cur_pages;
308 unsigned long ra_max_pages;
309 unsigned long ra_max_pages_per_file;
310 unsigned long ra_max_read_ahead_whole_pages;
313 /* ra_io_arg will be filled in the beginning of ll_readahead with
314 * ras_lock, then the following ll_read_ahead_pages will read RA
315 * pages according to this arg, all the items in this structure are
316 * counted by page index.
319 unsigned long ria_start; /* start offset of read-ahead*/
320 unsigned long ria_end; /* end offset of read-ahead*/
321 unsigned long ria_reserved; /* reserved pages for read-ahead */
322 unsigned long ria_end_min; /* minimum end to cover current read */
323 bool ria_eof; /* reach end of file */
324 /* If stride read pattern is detected, ria_stoff means where
325 * stride read is started. Note: for normal read-ahead, the
326 * value here is meaningless, and also it will not be accessed
329 /* ria_length and ria_pages are the length and pages length in the
330 * stride I/O mode. And they will also be used to check whether
331 * it is stride I/O read-ahead in the read-ahead pages
333 unsigned long ria_length;
334 unsigned long ria_pages;
337 /* LL_HIST_MAX=32 causes an overflow */
338 #define LL_HIST_MAX 28
339 #define LL_HIST_START 12 /* buckets start at 2^12 = 4k */
340 #define LL_PROCESS_HIST_MAX 10
341 struct per_process_info {
343 struct obd_histogram pp_r_hist;
344 struct obd_histogram pp_w_hist;
347 /* pp_extents[LL_PROCESS_HIST_MAX] will hold the combined process info */
348 struct ll_rw_extents_info {
349 struct per_process_info pp_extents[LL_PROCESS_HIST_MAX + 1];
352 #define LL_OFFSET_HIST_MAX 100
353 struct ll_rw_process_info {
356 loff_t rw_range_start;
358 loff_t rw_last_file_pos;
360 size_t rw_smallest_extent;
361 size_t rw_largest_extent;
362 struct ll_file_data *rw_last_file;
365 enum stats_track_type {
366 STATS_TRACK_ALL = 0, /* track all processes */
367 STATS_TRACK_PID, /* track process with this pid */
368 STATS_TRACK_PPID, /* track processes with this ppid */
369 STATS_TRACK_GID, /* track processes with this gid */
373 /* flags for sbi->ll_flags */
374 #define LL_SBI_NOLCK 0x01 /* DLM locking disabled (directio-only) */
375 #define LL_SBI_CHECKSUM 0x02 /* checksum each page as it's written */
376 #define LL_SBI_FLOCK 0x04
377 #define LL_SBI_USER_XATTR 0x08 /* support user xattr */
378 #define LL_SBI_ACL 0x10 /* support ACL */
379 /* LL_SBI_RMT_CLIENT 0x40 remote client */
380 #define LL_SBI_MDS_CAPA 0x80 /* support mds capa, obsolete */
381 #define LL_SBI_OSS_CAPA 0x100 /* support oss capa, obsolete */
382 #define LL_SBI_LOCALFLOCK 0x200 /* Local flocks support by kernel */
383 #define LL_SBI_LRU_RESIZE 0x400 /* lru resize support */
384 #define LL_SBI_LAZYSTATFS 0x800 /* lazystatfs mount option */
385 /* LL_SBI_SOM_PREVIEW 0x1000 SOM preview mount option, obsolete */
386 #define LL_SBI_32BIT_API 0x2000 /* generate 32 bit inodes. */
387 #define LL_SBI_64BIT_HASH 0x4000 /* support 64-bits dir hash/offset */
388 #define LL_SBI_AGL_ENABLED 0x8000 /* enable agl */
389 #define LL_SBI_VERBOSE 0x10000 /* verbose mount/umount */
390 #define LL_SBI_LAYOUT_LOCK 0x20000 /* layout lock support */
391 #define LL_SBI_USER_FID2PATH 0x40000 /* allow fid2path by unprivileged users */
392 #define LL_SBI_XATTR_CACHE 0x80000 /* support for xattr cache */
393 #define LL_SBI_NOROOTSQUASH 0x100000 /* do not apply root squash */
394 #define LL_SBI_ALWAYS_PING 0x200000 /* always ping even if server
397 #define LL_SBI_FLAGS { \
423 * This is embedded into llite super-blocks to keep track of connect
424 * flags (capabilities) supported by all imports given mount is
427 struct lustre_client_ocd {
429 * This is conjunction of connect_flags across all imports
430 * (LOVs) this mount is connected to. This field is updated by
431 * cl_ocd_update() under ->lco_lock.
434 struct mutex lco_lock;
435 struct obd_export *lco_md_exp;
436 struct obd_export *lco_dt_exp;
440 /* this protects pglist and ra_info. It isn't safe to
441 * grab from interrupt contexts
444 spinlock_t ll_pp_extent_lock; /* pp_extent entry*/
445 spinlock_t ll_process_lock; /* ll_rw_process_info */
446 struct obd_uuid ll_sb_uuid;
447 struct obd_export *ll_md_exp;
448 struct obd_export *ll_dt_exp;
449 struct dentry *ll_debugfs_entry;
450 struct lu_fid ll_root_fid; /* root object fid */
453 unsigned int ll_umounting:1,
454 ll_xattr_cache_enabled:1,
455 ll_client_common_fill_super_succeeded:1;
457 struct lustre_client_ocd ll_lco;
459 struct lprocfs_stats *ll_stats; /* lprocfs stats counter */
462 * Used to track "unstable" pages on a client, and maintain a
463 * LRU list of clean pages. An "unstable" page is defined as
464 * any page which is sent to a server as part of a bulk request,
465 * but is uncommitted to stable storage.
467 struct cl_client_cache *ll_cache;
469 struct lprocfs_stats *ll_ra_stats;
471 struct ll_ra_info ll_ra_info;
472 unsigned int ll_namelen;
473 const struct file_operations *ll_fop;
475 unsigned int ll_md_brw_pages; /* readdir pages per RPC */
477 struct lu_site *ll_site;
478 struct cl_device *ll_cl;
480 struct ll_rw_extents_info ll_rw_extents_info;
481 int ll_extent_process_count;
482 struct ll_rw_process_info ll_rw_process_info[LL_PROCESS_HIST_MAX];
483 unsigned int ll_offset_process_count;
484 struct ll_rw_process_info ll_rw_offset_info[LL_OFFSET_HIST_MAX];
485 unsigned int ll_rw_offset_entry_count;
486 int ll_stats_track_id;
487 enum stats_track_type ll_stats_track_type;
490 /* metadata stat-ahead */
491 unsigned int ll_sa_max; /* max statahead RPCs */
492 atomic_t ll_sa_total; /* statahead thread started
495 atomic_t ll_sa_wrong; /* statahead thread stopped for
498 atomic_t ll_sa_running; /* running statahead thread
501 atomic_t ll_agl_total; /* AGL thread started count */
503 dev_t ll_sdev_orig; /* save s_dev before assign for
507 struct root_squash_info ll_squash;
510 __kernel_fsid_t ll_fsid;
511 struct kobject ll_kobj; /* sysfs object */
512 struct super_block *ll_sb; /* struct super_block (for sysfs code)*/
513 struct completion ll_kobj_unregister;
517 * per file-descriptor read-ahead data.
519 struct ll_readahead_state {
522 * index of the last page that read(2) needed and that wasn't in the
523 * cache. Used by ras_update() to detect seeks.
525 * XXX nikita: if access seeks into cached region, Lustre doesn't see
528 unsigned long ras_last_readpage;
530 * number of pages read after last read-ahead window reset. As window
531 * is reset on each seek, this is effectively a number of consecutive
532 * accesses. Maybe ->ras_accessed_in_window is better name.
534 * XXX nikita: window is also reset (by ras_update()) when Lustre
535 * believes that memory pressure evicts read-ahead pages. In that
536 * case, it probably doesn't make sense to expand window to
537 * PTLRPC_MAX_BRW_PAGES on the third access.
539 unsigned long ras_consecutive_pages;
541 * number of read requests after the last read-ahead window reset
542 * As window is reset on each seek, this is effectively the number
543 * on consecutive read request and is used to trigger read-ahead.
545 unsigned long ras_consecutive_requests;
547 * Parameters of current read-ahead window. Handled by
548 * ras_update(). On the initial access to the file or after a seek,
549 * window is reset to 0. After 3 consecutive accesses, window is
550 * expanded to PTLRPC_MAX_BRW_PAGES. Afterwards, window is enlarged by
551 * PTLRPC_MAX_BRW_PAGES chunks up to ->ra_max_pages.
553 unsigned long ras_window_start, ras_window_len;
555 * Optimal RPC size. It decides how many pages will be sent
556 * for each read-ahead.
558 unsigned long ras_rpc_size;
560 * Where next read-ahead should start at. This lies within read-ahead
561 * window. Read-ahead window is read in pieces rather than at once
562 * because: 1. lustre limits total number of pages under read-ahead by
563 * ->ra_max_pages (see ll_ra_count_get()), 2. client cannot read pages
564 * not covered by DLM lock.
566 unsigned long ras_next_readahead;
568 * Total number of ll_file_read requests issued, reads originating
569 * due to mmap are not counted in this total. This value is used to
570 * trigger full file read-ahead after multiple reads to a small file.
572 unsigned long ras_requests;
574 * Page index with respect to the current request, these value
575 * will not be accurate when dealing with reads issued via mmap.
577 unsigned long ras_request_index;
579 * The following 3 items are used for detecting the stride I/O
581 * In stride I/O mode,
582 * ...............|-----data-----|****gap*****|--------|******|....
583 * offset |-stride_pages-|-stride_gap-|
584 * ras_stride_offset = offset;
585 * ras_stride_length = stride_pages + stride_gap;
586 * ras_stride_pages = stride_pages;
587 * Note: all these three items are counted by pages.
589 unsigned long ras_stride_length;
590 unsigned long ras_stride_pages;
591 pgoff_t ras_stride_offset;
593 * number of consecutive stride request count, and it is similar as
594 * ras_consecutive_requests, but used for stride I/O mode.
595 * Note: only more than 2 consecutive stride request are detected,
596 * stride read-ahead will be enable
598 unsigned long ras_consecutive_stride_requests;
601 extern struct kmem_cache *ll_file_data_slab;
602 struct lustre_handle;
603 struct ll_file_data {
604 struct ll_readahead_state fd_ras;
605 struct ll_grouplock fd_grouplock;
609 /* openhandle if lease exists for this file.
610 * Borrow lli->lli_och_mutex to protect assignment
612 struct obd_client_handle *fd_lease_och;
613 struct obd_client_handle *fd_och;
614 struct file *fd_file;
615 /* Indicate whether need to report failure when close.
616 * true: failure is known, not report again.
617 * false: unknown failure, should report.
619 bool fd_write_failed;
620 rwlock_t fd_lock; /* protect lcc list */
621 struct list_head fd_lccs; /* list of ll_cl_context */
624 extern struct dentry *llite_root;
625 extern struct kset *llite_kset;
627 static inline struct inode *ll_info2i(struct ll_inode_info *lli)
629 return &lli->lli_vfs_inode;
632 __u32 ll_i2suppgid(struct inode *i);
633 void ll_i2gids(__u32 *suppgids, struct inode *i1, struct inode *i2);
635 static inline int ll_need_32bit_api(struct ll_sb_info *sbi)
637 #if BITS_PER_LONG == 32
639 #elif defined(CONFIG_COMPAT)
640 return unlikely(in_compat_syscall() || (sbi->ll_flags & LL_SBI_32BIT_API));
642 return unlikely(sbi->ll_flags & LL_SBI_32BIT_API);
646 void ll_ras_enter(struct file *f);
648 /* llite/lcommon_misc.c */
649 int cl_init_ea_size(struct obd_export *md_exp, struct obd_export *dt_exp);
650 int cl_ocd_update(struct obd_device *host,
651 struct obd_device *watched,
652 enum obd_notify_event ev, void *owner, void *data);
653 int cl_get_grouplock(struct cl_object *obj, unsigned long gid, int nonblock,
654 struct ll_grouplock *cg);
655 void cl_put_grouplock(struct ll_grouplock *cg);
657 /* llite/lproc_llite.c */
658 int ldebugfs_register_mountpoint(struct dentry *parent,
659 struct super_block *sb, char *osc, char *mdc);
660 void ldebugfs_unregister_mountpoint(struct ll_sb_info *sbi);
661 void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count);
662 void lprocfs_llite_init_vars(struct lprocfs_static_vars *lvars);
663 void ll_rw_stats_tally(struct ll_sb_info *sbi, pid_t pid,
664 struct ll_file_data *file, loff_t pos,
665 size_t count, int rw);
669 LPROC_LL_DIRTY_MISSES,
671 LPROC_LL_WRITE_BYTES,
694 LPROC_LL_ALLOC_INODE,
697 LPROC_LL_GETXATTR_HITS,
699 LPROC_LL_REMOVEXATTR,
701 LPROC_LL_FILE_OPCODES
705 extern const struct file_operations ll_dir_operations;
706 extern const struct inode_operations ll_dir_inode_operations;
707 int ll_dir_read(struct inode *inode, __u64 *ppos, struct md_op_data *op_data,
708 struct dir_context *ctx);
709 int ll_get_mdt_idx(struct inode *inode);
710 int ll_get_mdt_idx_by_fid(struct ll_sb_info *sbi, const struct lu_fid *fid);
711 struct page *ll_get_dir_page(struct inode *dir, struct md_op_data *op_data,
713 void ll_release_page(struct inode *inode, struct page *page, bool remove);
716 extern const struct inode_operations ll_special_inode_operations;
718 struct inode *ll_iget(struct super_block *sb, ino_t hash,
719 struct lustre_md *lic);
720 int ll_test_inode_by_fid(struct inode *inode, void *opaque);
721 int ll_md_blocking_ast(struct ldlm_lock *lock, struct ldlm_lock_desc *desc,
722 void *data, int flag);
723 struct dentry *ll_splice_alias(struct inode *inode, struct dentry *de);
724 void ll_update_times(struct ptlrpc_request *request, struct inode *inode);
727 int ll_writepage(struct page *page, struct writeback_control *wbc);
728 int ll_writepages(struct address_space *mapping, struct writeback_control *wbc);
729 int ll_readpage(struct file *file, struct page *page);
730 void ll_readahead_init(struct inode *inode, struct ll_readahead_state *ras);
731 int vvp_io_write_commit(const struct lu_env *env, struct cl_io *io);
732 struct ll_cl_context *ll_cl_find(struct file *file);
733 void ll_cl_add(struct file *file, const struct lu_env *env, struct cl_io *io);
734 void ll_cl_remove(struct file *file, const struct lu_env *env);
736 extern const struct address_space_operations ll_aops;
739 extern const struct file_operations ll_file_operations;
740 extern const struct file_operations ll_file_operations_flock;
741 extern const struct file_operations ll_file_operations_noflock;
742 extern const struct inode_operations ll_file_inode_operations;
743 int ll_have_md_lock(struct inode *inode, __u64 *bits,
744 enum ldlm_mode l_req_mode);
745 enum ldlm_mode ll_take_md_lock(struct inode *inode, __u64 bits,
746 struct lustre_handle *lockh, __u64 flags,
747 enum ldlm_mode mode);
748 int ll_file_open(struct inode *inode, struct file *file);
749 int ll_file_release(struct inode *inode, struct file *file);
750 int ll_release_openhandle(struct inode *inode, struct lookup_intent *it);
751 int ll_md_real_close(struct inode *inode, fmode_t fmode);
752 int ll_getattr(const struct path *path, struct kstat *stat,
753 u32 request_mask, unsigned int flags);
754 struct posix_acl *ll_get_acl(struct inode *inode, int type);
755 int ll_migrate(struct inode *parent, struct file *file, int mdtidx,
756 const char *name, int namelen);
757 int ll_get_fid_by_name(struct inode *parent, const char *name,
758 int namelen, struct lu_fid *fid, struct inode **inode);
759 int ll_inode_permission(struct inode *inode, int mask);
761 int ll_lov_setstripe_ea_info(struct inode *inode, struct dentry *dentry,
762 __u64 flags, struct lov_user_md *lum,
764 int ll_lov_getstripe_ea_info(struct inode *inode, const char *filename,
765 struct lov_mds_md **lmm, int *lmm_size,
766 struct ptlrpc_request **request);
767 int ll_dir_setstripe(struct inode *inode, struct lov_user_md *lump,
769 int ll_dir_getstripe(struct inode *inode, void **lmmp, int *lmm_size,
770 struct ptlrpc_request **request, u64 valid);
771 int ll_fsync(struct file *file, loff_t start, loff_t end, int data);
772 int ll_merge_attr(const struct lu_env *env, struct inode *inode);
773 int ll_fid2path(struct inode *inode, void __user *arg);
774 int ll_data_version(struct inode *inode, __u64 *data_version, int flags);
775 int ll_hsm_release(struct inode *inode);
776 int ll_hsm_state_set(struct inode *inode, struct hsm_state_set *hss);
780 extern const struct dentry_operations ll_d_ops;
781 void ll_intent_drop_lock(struct lookup_intent *it);
782 void ll_intent_release(struct lookup_intent *it);
783 void ll_invalidate_aliases(struct inode *inode);
784 void ll_lookup_finish_locks(struct lookup_intent *it, struct inode *inode);
785 int ll_revalidate_it_finish(struct ptlrpc_request *request,
786 struct lookup_intent *it, struct inode *inode);
788 /* llite/llite_lib.c */
789 extern struct super_operations lustre_super_operations;
791 void ll_lli_init(struct ll_inode_info *lli);
792 int ll_fill_super(struct super_block *sb, struct vfsmount *mnt);
793 void ll_put_super(struct super_block *sb);
794 void ll_kill_super(struct super_block *sb);
795 struct inode *ll_inode_from_resource_lock(struct ldlm_lock *lock);
796 void ll_dir_clear_lsm_md(struct inode *inode);
797 void ll_clear_inode(struct inode *inode);
798 int ll_setattr_raw(struct dentry *dentry, struct iattr *attr, bool hsm_import);
799 int ll_setattr(struct dentry *de, struct iattr *attr);
800 int ll_statfs(struct dentry *de, struct kstatfs *sfs);
801 int ll_statfs_internal(struct super_block *sb, struct obd_statfs *osfs,
802 __u64 max_age, __u32 flags);
803 int ll_update_inode(struct inode *inode, struct lustre_md *md);
804 int ll_read_inode2(struct inode *inode, void *opaque);
805 void ll_delete_inode(struct inode *inode);
806 int ll_iocontrol(struct inode *inode, struct file *file,
807 unsigned int cmd, unsigned long arg);
808 int ll_flush_ctx(struct inode *inode);
809 void ll_umount_begin(struct super_block *sb);
810 int ll_remount_fs(struct super_block *sb, int *flags, char *data);
811 int ll_show_options(struct seq_file *seq, struct dentry *dentry);
812 void ll_dirty_page_discard_warn(struct page *page, int ioret);
813 int ll_prep_inode(struct inode **inode, struct ptlrpc_request *req,
814 struct super_block *sb, struct lookup_intent *it);
815 int ll_obd_statfs(struct inode *inode, void __user *arg);
816 int ll_get_max_mdsize(struct ll_sb_info *sbi, int *max_mdsize);
817 int ll_get_default_mdsize(struct ll_sb_info *sbi, int *default_mdsize);
818 int ll_set_default_mdsize(struct ll_sb_info *sbi, int default_mdsize);
819 int ll_process_config(struct lustre_cfg *lcfg);
822 LUSTRE_OPC_MKDIR = 0,
823 LUSTRE_OPC_SYMLINK = 1,
824 LUSTRE_OPC_MKNOD = 2,
825 LUSTRE_OPC_CREATE = 3,
829 struct md_op_data *ll_prep_md_op_data(struct md_op_data *op_data,
830 struct inode *i1, struct inode *i2,
831 const char *name, size_t namelen,
832 u32 mode, __u32 opc, void *data);
833 void ll_finish_md_op_data(struct md_op_data *op_data);
834 int ll_get_obd_name(struct inode *inode, unsigned int cmd, unsigned long arg);
835 char *ll_get_fsname(struct super_block *sb, char *buf, int buflen);
836 void ll_compute_rootsquash_state(struct ll_sb_info *sbi);
837 void ll_open_cleanup(struct super_block *sb, struct ptlrpc_request *open_req);
838 ssize_t ll_copy_user_md(const struct lov_user_md __user *md,
839 struct lov_user_md **kbuf);
841 /* Compute expected user md size when passing in a md from user space */
842 static inline ssize_t ll_lov_user_md_size(const struct lov_user_md *lum)
844 switch (lum->lmm_magic) {
845 case LOV_USER_MAGIC_V1:
846 return sizeof(struct lov_user_md_v1);
847 case LOV_USER_MAGIC_V3:
848 return sizeof(struct lov_user_md_v3);
849 case LOV_USER_MAGIC_SPECIFIC:
850 if (lum->lmm_stripe_count > LOV_MAX_STRIPE_COUNT)
853 return lov_user_md_size(lum->lmm_stripe_count,
854 LOV_USER_MAGIC_SPECIFIC);
859 /* llite/llite_nfs.c */
860 extern const struct export_operations lustre_export_operations;
861 __u32 get_uuid2int(const char *name, int len);
862 void get_uuid2fsid(const char *name, int len, __kernel_fsid_t *fsid);
863 struct inode *search_inode_for_lustre(struct super_block *sb,
864 const struct lu_fid *fid);
865 int ll_dir_get_parent_fid(struct inode *dir, struct lu_fid *parent_fid);
867 /* llite/symlink.c */
868 extern const struct inode_operations ll_fast_symlink_inode_operations;
871 * IO arguments for various VFS I/O interfaces.
877 struct kiocb *via_iocb;
878 struct iov_iter *via_iter;
883 struct ll_cl_context {
884 struct list_head lcc_list;
886 const struct lu_env *lcc_env;
887 struct cl_io *lcc_io;
888 struct cl_page *lcc_page;
891 struct ll_thread_info {
892 struct vvp_io_args lti_args;
893 struct ra_io_arg lti_ria;
894 struct ll_cl_context lti_io_ctx;
897 extern struct lu_context_key ll_thread_key;
898 static inline struct ll_thread_info *ll_env_info(const struct lu_env *env)
900 struct ll_thread_info *lti;
902 lti = lu_context_key_get(&env->le_ctx, &ll_thread_key);
907 static inline struct vvp_io_args *ll_env_args(const struct lu_env *env)
909 return &ll_env_info(env)->lti_args;
912 /* llite/llite_mmap.c */
914 int ll_teardown_mmaps(struct address_space *mapping, __u64 first, __u64 last);
915 int ll_file_mmap(struct file *file, struct vm_area_struct *vma);
916 void policy_from_vma(union ldlm_policy_data *policy, struct vm_area_struct *vma,
917 unsigned long addr, size_t count);
918 struct vm_area_struct *our_vma(struct mm_struct *mm, unsigned long addr,
921 static inline void ll_invalidate_page(struct page *vmpage)
923 struct address_space *mapping = vmpage->mapping;
924 loff_t offset = vmpage->index << PAGE_SHIFT;
926 LASSERT(PageLocked(vmpage));
931 * truncate_complete_page() calls
932 * a_ops->invalidatepage()->cl_page_delete()->vvp_page_delete().
934 ll_teardown_mmaps(mapping, offset, offset + PAGE_SIZE);
935 truncate_complete_page(mapping, vmpage);
938 #define ll_s2sbi(sb) (s2lsi(sb)->lsi_llsbi)
940 /* don't need an addref as the sb_info should be holding one */
941 static inline struct obd_export *ll_s2dtexp(struct super_block *sb)
943 return ll_s2sbi(sb)->ll_dt_exp;
946 /* don't need an addref as the sb_info should be holding one */
947 static inline struct obd_export *ll_s2mdexp(struct super_block *sb)
949 return ll_s2sbi(sb)->ll_md_exp;
952 static inline struct client_obd *sbi2mdc(struct ll_sb_info *sbi)
954 struct obd_device *obd = sbi->ll_md_exp->exp_obd;
961 /* FIXME: replace the name of this with LL_SB to conform to kernel stuff */
962 static inline struct ll_sb_info *ll_i2sbi(struct inode *inode)
964 return ll_s2sbi(inode->i_sb);
967 static inline struct obd_export *ll_i2dtexp(struct inode *inode)
969 return ll_s2dtexp(inode->i_sb);
972 static inline struct obd_export *ll_i2mdexp(struct inode *inode)
974 return ll_s2mdexp(inode->i_sb);
977 static inline struct lu_fid *ll_inode2fid(struct inode *inode)
982 fid = &ll_i2info(inode)->lli_fid;
987 static inline loff_t ll_file_maxbytes(struct inode *inode)
989 struct cl_object *obj = ll_i2info(inode)->lli_clob;
992 return MAX_LFS_FILESIZE;
994 return min_t(loff_t, cl_object_maxbytes(obj), MAX_LFS_FILESIZE);
998 extern const struct xattr_handler *ll_xattr_handlers[];
1000 #define XATTR_USER_T 1
1001 #define XATTR_TRUSTED_T 2
1002 #define XATTR_SECURITY_T 3
1003 #define XATTR_ACL_ACCESS_T 4
1004 #define XATTR_ACL_DEFAULT_T 5
1005 #define XATTR_LUSTRE_T 6
1006 #define XATTR_OTHER_T 7
1008 ssize_t ll_listxattr(struct dentry *dentry, char *buffer, size_t size);
1009 int ll_xattr_list(struct inode *inode, const char *name, int type,
1010 void *buffer, size_t size, __u64 valid);
1011 const struct xattr_handler *get_xattr_type(const char *name);
1014 * Common IO arguments for various VFS I/O interfaces.
1016 int cl_sb_init(struct super_block *sb);
1017 int cl_sb_fini(struct super_block *sb);
1019 enum ras_update_flags {
1023 void ll_ra_count_put(struct ll_sb_info *sbi, unsigned long len);
1024 void ll_ra_stats_inc(struct inode *inode, enum ra_stat which);
1027 #define LL_SA_RPC_MIN 2
1028 #define LL_SA_RPC_DEF 32
1029 #define LL_SA_RPC_MAX 8192
1031 #define LL_SA_CACHE_BIT 5
1032 #define LL_SA_CACHE_SIZE (1 << LL_SA_CACHE_BIT)
1033 #define LL_SA_CACHE_MASK (LL_SA_CACHE_SIZE - 1)
1035 /* per inode struct, for dir only */
1036 struct ll_statahead_info {
1037 struct dentry *sai_dentry;
1038 atomic_t sai_refcount; /* when access this struct, hold
1041 unsigned int sai_max; /* max ahead of lookup */
1042 __u64 sai_sent; /* stat requests sent count */
1043 __u64 sai_replied; /* stat requests which received
1046 __u64 sai_index; /* index of statahead entry */
1047 __u64 sai_index_wait; /* index of entry which is the
1048 * caller is waiting for
1050 __u64 sai_hit; /* hit count */
1051 __u64 sai_miss; /* miss count:
1052 * for "ls -al" case, it includes
1053 * hidden dentry miss;
1054 * for "ls -l" case, it does not
1055 * include hidden dentry miss.
1056 * "sai_miss_hidden" is used for
1059 unsigned int sai_consecutive_miss; /* consecutive miss */
1060 unsigned int sai_miss_hidden;/* "ls -al", but first dentry
1061 * is not a hidden one
1063 unsigned int sai_skip_hidden;/* skipped hidden dentry count */
1064 unsigned int sai_ls_all:1, /* "ls -al", do stat-ahead for
1067 sai_agl_valid:1,/* AGL is valid for the dir */
1068 sai_in_readpage:1;/* statahead is in readdir() */
1069 wait_queue_head_t sai_waitq; /* stat-ahead wait queue */
1070 struct ptlrpc_thread sai_thread; /* stat-ahead thread */
1071 struct ptlrpc_thread sai_agl_thread; /* AGL thread */
1072 struct list_head sai_interim_entries; /* entries which got async
1073 * stat reply, but not
1076 struct list_head sai_entries; /* completed entries */
1077 struct list_head sai_agls; /* AGLs to be sent */
1078 struct list_head sai_cache[LL_SA_CACHE_SIZE];
1079 spinlock_t sai_cache_lock[LL_SA_CACHE_SIZE];
1080 atomic_t sai_cache_count; /* entry count in cache */
1083 int ll_statahead(struct inode *dir, struct dentry **dentry, bool unplug);
1084 void ll_authorize_statahead(struct inode *dir, void *key);
1085 void ll_deauthorize_statahead(struct inode *dir, void *key);
1087 blkcnt_t dirty_cnt(struct inode *inode);
1089 int cl_glimpse_size0(struct inode *inode, int agl);
1090 int cl_glimpse_lock(const struct lu_env *env, struct cl_io *io,
1091 struct inode *inode, struct cl_object *clob, int agl);
1093 static inline int cl_glimpse_size(struct inode *inode)
1095 return cl_glimpse_size0(inode, 0);
1098 static inline int cl_agl(struct inode *inode)
1100 return cl_glimpse_size0(inode, 1);
1103 static inline int ll_glimpse_size(struct inode *inode)
1105 struct ll_inode_info *lli = ll_i2info(inode);
1108 down_read(&lli->lli_glimpse_sem);
1109 rc = cl_glimpse_size(inode);
1110 lli->lli_glimpse_time = cfs_time_current();
1111 up_read(&lli->lli_glimpse_sem);
1116 * dentry may statahead when statahead is enabled and current process has opened
1117 * parent directory, and this dentry hasn't accessed statahead cache before
1120 dentry_may_statahead(struct inode *dir, struct dentry *dentry)
1122 struct ll_inode_info *lli;
1123 struct ll_dentry_data *ldd;
1125 if (ll_i2sbi(dir)->ll_sa_max == 0)
1128 lli = ll_i2info(dir);
1131 * statahead is not allowed for this dir, there may be three causes:
1132 * 1. dir is not opened.
1133 * 2. statahead hit ratio is too low.
1134 * 3. previous stat started statahead thread failed.
1136 if (!lli->lli_sa_enabled)
1139 /* not the same process, don't statahead */
1140 if (lli->lli_opendir_pid != current_pid())
1144 * When stating a dentry, kernel may trigger 'revalidate' or 'lookup'
1145 * multiple times, eg. for 'getattr', 'getxattr' and etc.
1146 * For patchless client, lookup intent is not accurate, which may
1147 * misguide statahead. For example:
1148 * The 'revalidate' call for 'getattr' and 'getxattr' of a dentry will
1149 * have the same intent -- IT_GETATTR, while one dentry should access
1150 * statahead cache once, otherwise statahead windows is messed up.
1151 * The solution is as following:
1152 * Assign 'lld_sa_generation' with 'lli_sa_generation' when a dentry
1153 * IT_GETATTR for the first time, and subsequent IT_GETATTR will
1154 * bypass interacting with statahead cache by checking
1155 * 'lld_sa_generation == lli->lli_sa_generation'.
1157 ldd = ll_d2d(dentry);
1158 if (ldd->lld_sa_generation == lli->lli_sa_generation)
1164 /* llite ioctl register support routine */
1170 #define LLIOC_MAX_CMD 256
1173 * Rules to write a callback function:
1176 * @magic: Dynamic ioctl call routine will feed this value with the pointer
1177 * returned to ll_iocontrol_register. Callback functions should use this
1178 * data to check the potential collasion of ioctl cmd. If collasion is
1179 * found, callback function should return LLIOC_CONT.
1180 * @rcp: The result of ioctl command.
1183 * If @magic matches the pointer returned by ll_iocontrol_data, the
1184 * callback should return LLIOC_STOP; return LLIOC_STOP otherwise.
1186 typedef enum llioc_iter (*llioc_callback_t)(struct inode *inode,
1187 struct file *file, unsigned int cmd, unsigned long arg,
1188 void *magic, int *rcp);
1190 /* export functions */
1191 /* Register ioctl block dynamatically for a regular file.
1193 * @cmd: the array of ioctl command set
1194 * @count: number of commands in the @cmd
1195 * @cb: callback function, it will be called if an ioctl command is found to
1196 * belong to the command list @cmd.
1199 * A magic pointer will be returned if success;
1200 * otherwise, NULL will be returned.
1202 void *ll_iocontrol_register(llioc_callback_t cb, int count, unsigned int *cmd);
1203 void ll_iocontrol_unregister(void *magic);
1205 int cl_sync_file_range(struct inode *inode, loff_t start, loff_t end,
1206 enum cl_fsync_mode mode, int ignore_layout);
1208 /** direct write pages */
1209 struct ll_dio_pages {
1210 /** page array to be written. we don't support
1211 * partial pages except the last one.
1213 struct page **ldp_pages;
1214 /* offset of each page */
1215 loff_t *ldp_offsets;
1216 /** if ldp_offsets is NULL, it means a sequential
1217 * pages to be written, then this is the file offset
1218 * of the first page.
1220 loff_t ldp_start_offset;
1221 /** how many bytes are to be written. */
1223 /** # of pages in the array. */
1227 ssize_t ll_direct_rw_pages(const struct lu_env *env, struct cl_io *io,
1228 int rw, struct inode *inode,
1229 struct ll_dio_pages *pv);
1231 static inline int ll_file_nolock(const struct file *file)
1233 struct ll_file_data *fd = LUSTRE_FPRIVATE(file);
1234 struct inode *inode = file_inode(file);
1236 return ((fd->fd_flags & LL_FILE_IGNORE_LOCK) ||
1237 (ll_i2sbi(inode)->ll_flags & LL_SBI_NOLCK));
1240 static inline void ll_set_lock_data(struct obd_export *exp, struct inode *inode,
1241 struct lookup_intent *it, __u64 *bits)
1243 if (!it->it_lock_set) {
1244 struct lustre_handle handle;
1246 /* If this inode is a remote object, it will get two
1247 * separate locks in different namespaces, Master MDT,
1248 * where the name entry is, will grant LOOKUP lock,
1249 * remote MDT, where the object is, will grant
1250 * UPDATE|PERM lock. The inode will be attached to both
1251 * LOOKUP and PERM locks, so revoking either locks will
1252 * case the dcache being cleared
1254 if (it->it_remote_lock_mode) {
1255 handle.cookie = it->it_remote_lock_handle;
1256 CDEBUG(D_DLMTRACE, "setting l_data to inode " DFID "%p for remote lock %#llx\n",
1257 PFID(ll_inode2fid(inode)), inode,
1259 md_set_lock_data(exp, &handle, inode, NULL);
1262 handle.cookie = it->it_lock_handle;
1264 CDEBUG(D_DLMTRACE, "setting l_data to inode " DFID "%p for lock %#llx\n",
1265 PFID(ll_inode2fid(inode)), inode, handle.cookie);
1267 md_set_lock_data(exp, &handle, inode, &it->it_lock_bits);
1268 it->it_lock_set = 1;
1272 *bits = it->it_lock_bits;
1275 static inline int d_lustre_invalid(const struct dentry *dentry)
1277 return ll_d2d(dentry)->lld_invalid;
1281 * Mark dentry INVALID, if dentry refcount is zero (this is normally case for
1282 * ll_md_blocking_ast), unhash this dentry, and let dcache to reclaim it later;
1283 * else dput() of the last refcount will unhash this dentry and kill it.
1285 static inline void d_lustre_invalidate(struct dentry *dentry, int nested)
1287 CDEBUG(D_DENTRY, "invalidate dentry %pd (%p) parent %p inode %p refc %d\n",
1289 dentry->d_parent, d_inode(dentry), d_count(dentry));
1291 spin_lock_nested(&dentry->d_lock,
1292 nested ? DENTRY_D_LOCK_NESTED : DENTRY_D_LOCK_NORMAL);
1293 ll_d2d(dentry)->lld_invalid = 1;
1295 * We should be careful about dentries created by d_obtain_alias().
1296 * These dentries are not put in the dentry tree, instead they are
1297 * linked to sb->s_anon through dentry->d_hash.
1298 * shrink_dcache_for_umount() shrinks the tree and sb->s_anon list.
1299 * If we unhashed such a dentry, unmount would not be able to find
1300 * it and busy inodes would be reported.
1302 if (d_count(dentry) == 0 && !(dentry->d_flags & DCACHE_DISCONNECTED))
1304 spin_unlock(&dentry->d_lock);
1307 static inline void d_lustre_revalidate(struct dentry *dentry)
1309 spin_lock(&dentry->d_lock);
1310 LASSERT(ll_d2d(dentry));
1311 ll_d2d(dentry)->lld_invalid = 0;
1312 spin_unlock(&dentry->d_lock);
1315 int ll_layout_conf(struct inode *inode, const struct cl_object_conf *conf);
1316 int ll_layout_refresh(struct inode *inode, __u32 *gen);
1317 int ll_layout_restore(struct inode *inode, loff_t start, __u64 length);
1319 int ll_xattr_init(void);
1320 void ll_xattr_fini(void);
1322 int ll_page_sync_io(const struct lu_env *env, struct cl_io *io,
1323 struct cl_page *page, enum cl_req_type crt);
1325 int ll_getparent(struct file *file, struct getparent __user *arg);
1328 int cl_setattr_ost(struct cl_object *obj, const struct iattr *attr,
1329 unsigned int attr_flags);
1331 extern struct lu_env *cl_inode_fini_env;
1332 extern u16 cl_inode_fini_refcheck;
1334 int cl_file_inode_init(struct inode *inode, struct lustre_md *md);
1335 void cl_inode_fini(struct inode *inode);
1337 __u64 cl_fid_build_ino(const struct lu_fid *fid, int api32);
1338 __u32 cl_fid_build_gen(const struct lu_fid *fid);
1340 #endif /* LLITE_INTERNAL_H */