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.sun.com/software/products/lustre/docs/GPLv2.pdf
20 * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
21 * CA 95054 USA or visit www.sun.com if you need additional information or
27 * Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved.
28 * Use is subject to license terms.
30 * Copyright (c) 2011, 2014, Intel Corporation.
33 * This file is part of Lustre, http://www.lustre.org/
34 * Lustre is a trademark of Sun Microsystems, Inc.
37 #ifndef LLITE_INTERNAL_H
38 #define LLITE_INTERNAL_H
39 #include <lustre_debug.h>
40 #include <lustre_ver.h>
41 #include <lustre_disk.h> /* for s2sbi */
42 #include <lustre_eacl.h>
43 #include <lustre_linkea.h>
45 /* for struct cl_lock_descr and struct cl_io */
46 #include <cl_object.h>
47 #include <lustre_lmv.h>
48 #include <lustre_mdc.h>
49 #include <lustre_intent.h>
50 #include <linux/compat.h>
52 #include "vvp_internal.h"
53 #include "range_lock.h"
59 #ifndef 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")
77 #define LUSTRE_FPRIVATE(file) ((file)->private_data)
79 struct ll_dentry_data {
80 struct lookup_intent *lld_it;
81 unsigned int lld_sa_generation;
82 unsigned int lld_invalid:1;
83 struct rcu_head lld_rcu_head;
86 #define ll_d2d(de) ((struct ll_dentry_data*)((de)->d_fsdata))
88 #define LLI_INODE_MAGIC 0x111d0de5
89 #define LLI_INODE_DEAD 0xdeadd00d
91 /* remote client permission cache */
92 #define REMOTE_PERM_HASHSIZE 16
94 struct ll_getname_data {
95 #ifdef HAVE_DIR_CONTEXT
96 struct dir_context ctx;
98 char *lgd_name; /* points to a buffer with NAME_MAX+1 size */
99 struct lu_fid lgd_fid; /* target fid we are looking for */
100 int lgd_found; /* inode matched? */
103 /* llite setxid/access permission for user on remote client */
104 struct ll_remote_perm {
105 struct hlist_node lrp_list;
110 int lrp_access_perm; /* MAY_READ/WRITE/EXEC, this
111 is access permission with
112 lrp_fsuid/lrp_fsgid. */
116 /* MDS has an authority for the Size-on-MDS attributes. */
117 LLIF_MDS_SIZE_LOCK = (1 << 0),
118 /* Epoch close is postponed. */
119 LLIF_EPOCH_PENDING = (1 << 1),
120 /* DONE WRITING is allowed. */
121 LLIF_DONE_WRITING = (1 << 2),
122 /* Sizeon-on-MDS attributes are changed. An attribute update needs to
124 LLIF_SOM_DIRTY = (1 << 3),
125 /* File data is modified. */
126 LLIF_DATA_MODIFIED = (1 << 4),
127 /* File is being restored */
128 LLIF_FILE_RESTORING = (1 << 5),
129 /* Xattr cache is attached to the file */
130 LLIF_XATTR_CACHE = (1 << 6),
133 struct ll_inode_info {
134 __u32 lli_inode_magic;
139 struct posix_acl *lli_posix_acl;
141 struct hlist_head *lli_remote_perms;
142 struct mutex lli_rmtperm_mutex;
144 /* identifying fields for both metadata and data stacks. */
145 struct lu_fid lli_fid;
146 /* master inode fid for stripe directory */
147 struct lu_fid lli_pfid;
149 struct list_head lli_close_list;
150 struct list_head lli_oss_capas;
151 /* open count currently used by capability only, indicate whether
152 * capability needs renewal */
153 atomic_t lli_open_count;
154 struct obd_capa *lli_mds_capa;
155 cfs_time_t lli_rmtperm_time;
157 /* handle is to be sent to MDS later on done_writing and setattr.
158 * Open handle data are needed for the recovery to reconstruct
159 * the inode state on the MDS. XXX: recovery is not ready yet. */
160 struct obd_client_handle *lli_pending_och;
162 /* We need all three because every inode may be opened in different
164 struct obd_client_handle *lli_mds_read_och;
165 struct obd_client_handle *lli_mds_write_och;
166 struct obd_client_handle *lli_mds_exec_och;
167 __u64 lli_open_fd_read_count;
168 __u64 lli_open_fd_write_count;
169 __u64 lli_open_fd_exec_count;
170 /* Protects access to och pointers and their usage counters */
171 struct mutex lli_och_mutex;
173 struct inode lli_vfs_inode;
175 /* the most recent timestamps obtained from mds */
179 spinlock_t lli_agl_lock;
181 /* Try to make the d::member and f::member are aligned. Before using
182 * these members, make clear whether it is directory or not. */
186 /* serialize normal readdir and statahead-readdir. */
187 struct mutex lli_readdir_mutex;
189 /* metadata statahead */
190 /* since parent-child threads can share the same @file
191 * struct, "opendir_key" is the token when dir close for
192 * case of parent exit before child -- it is me should
193 * cleanup the dir readahead. */
194 void *lli_opendir_key;
195 struct ll_statahead_info *lli_sai;
196 /* protect statahead stuff. */
197 spinlock_t lli_sa_lock;
198 /* "opendir_pid" is the token when lookup/revalid
199 * -- I am the owner of dir statahead. */
200 pid_t lli_opendir_pid;
201 /* stat will try to access statahead entries or start
202 * statahead if this flag is set, and this flag will be
203 * set upon dir open, and cleared when dir is closed,
204 * statahead hit ratio is too low, or start statahead
206 unsigned int lli_sa_enabled:1;
207 /* generation for statahead */
208 unsigned int lli_sa_generation;
209 /* directory stripe information */
210 struct lmv_stripe_md *lli_lsm_md;
211 /* striped directory size */
212 loff_t lli_stripe_dir_size;
213 /* striped directory nlink */
214 __u64 lli_stripe_dir_nlink;
217 /* for non-directory */
219 struct mutex lli_size_mutex;
220 char *lli_symlink_name;
223 * struct rw_semaphore {
224 * signed long count; // align d.d_def_acl
225 * spinlock_t wait_lock; // align d.d_sa_lock
226 * struct list_head wait_list;
229 struct rw_semaphore lli_trunc_sem;
230 struct range_lock_tree lli_write_tree;
232 struct rw_semaphore lli_glimpse_sem;
233 cfs_time_t lli_glimpse_time;
234 struct list_head lli_agl_list;
237 /* for writepage() only to communicate to fsync */
241 * whenever a process try to read/write the file, the
242 * jobid of the process will be saved here, and it'll
243 * be packed into the write PRC when flush later.
245 * so the read/write statistics for jobid will not be
246 * accurate if the file is shared by different jobs.
248 char lli_jobid[LUSTRE_JOBID_SIZE];
252 /* XXX: For following frequent used members, although they maybe special
253 * used for non-directory object, it is some time-wasting to check
254 * whether the object is directory or not before using them. On the
255 * other hand, currently, sizeof(f) > sizeof(d), it cannot reduce
256 * the "ll_inode_info" size even if moving those members into u.f.
257 * So keep them out side.
259 * In the future, if more members are added only for directory,
260 * some of the following members can be moved into u.f.
263 struct cl_object *lli_clob;
265 /* mutex to request for layout lock exclusively. */
266 struct mutex lli_layout_mutex;
267 /* Layout version, protected by lli_layout_lock */
268 __u32 lli_layout_gen;
269 spinlock_t lli_layout_lock;
271 struct rw_semaphore lli_xattrs_list_rwsem;
272 struct mutex lli_xattrs_enq_lock;
273 struct list_head lli_xattrs; /* ll_xattr_entry->xe_list */
276 static inline __u32 ll_layout_version_get(struct ll_inode_info *lli)
280 spin_lock(&lli->lli_layout_lock);
281 gen = lli->lli_layout_gen;
282 spin_unlock(&lli->lli_layout_lock);
287 static inline void ll_layout_version_set(struct ll_inode_info *lli, __u32 gen)
289 spin_lock(&lli->lli_layout_lock);
290 lli->lli_layout_gen = gen;
291 spin_unlock(&lli->lli_layout_lock);
294 int ll_xattr_cache_destroy(struct inode *inode);
296 int ll_xattr_cache_get(struct inode *inode,
303 * Locking to guarantee consistency of non-atomic updates to long long i_size,
304 * consistency between file size and KMS.
306 * Implemented by ->lli_size_mutex and ->lsm_lock, nested in that order.
309 void ll_inode_size_lock(struct inode *inode);
310 void ll_inode_size_unlock(struct inode *inode);
312 // FIXME: replace the name of this with LL_I to conform to kernel stuff
313 // static inline struct ll_inode_info *LL_I(struct inode *inode)
314 static inline struct ll_inode_info *ll_i2info(struct inode *inode)
316 return container_of(inode, struct ll_inode_info, lli_vfs_inode);
319 /* default to about 40meg of readahead on a given system. That much tied
320 * up in 512k readahead requests serviced at 40ms each is about 1GB/s. */
321 #define SBI_DEFAULT_READAHEAD_MAX (40UL << (20 - PAGE_CACHE_SHIFT))
323 /* default to read-ahead full files smaller than 2MB on the second read */
324 #define SBI_DEFAULT_READAHEAD_WHOLE_MAX (2UL << (20 - PAGE_CACHE_SHIFT))
329 RA_STAT_DISTANT_READPAGE,
330 RA_STAT_MISS_IN_WINDOW,
331 RA_STAT_FAILED_GRAB_PAGE,
332 RA_STAT_FAILED_MATCH,
337 RA_STAT_MAX_IN_FLIGHT,
338 RA_STAT_WRONG_GRAB_PAGE,
339 RA_STAT_FAILED_REACH_END,
344 atomic_t ra_cur_pages;
345 unsigned long ra_max_pages;
346 unsigned long ra_max_pages_per_file;
347 unsigned long ra_max_read_ahead_whole_pages;
350 /* ra_io_arg will be filled in the beginning of ll_readahead with
351 * ras_lock, then the following ll_read_ahead_pages will read RA
352 * pages according to this arg, all the items in this structure are
353 * counted by page index.
356 unsigned long ria_start; /* start offset of read-ahead*/
357 unsigned long ria_end; /* end offset of read-ahead*/
358 /* If stride read pattern is detected, ria_stoff means where
359 * stride read is started. Note: for normal read-ahead, the
360 * value here is meaningless, and also it will not be accessed*/
362 /* ria_length and ria_pages are the length and pages length in the
363 * stride I/O mode. And they will also be used to check whether
364 * it is stride I/O read-ahead in the read-ahead pages*/
365 unsigned long ria_length;
366 unsigned long ria_pages;
369 /* LL_HIST_MAX=32 causes an overflow */
370 #define LL_HIST_MAX 28
371 #define LL_HIST_START 12 /* buckets start at 2^12 = 4k */
372 #define LL_PROCESS_HIST_MAX 10
373 struct per_process_info {
375 struct obd_histogram pp_r_hist;
376 struct obd_histogram pp_w_hist;
379 /* pp_extents[LL_PROCESS_HIST_MAX] will hold the combined process info */
380 struct ll_rw_extents_info {
381 struct per_process_info pp_extents[LL_PROCESS_HIST_MAX + 1];
384 #define LL_OFFSET_HIST_MAX 100
385 struct ll_rw_process_info {
388 loff_t rw_range_start;
390 loff_t rw_last_file_pos;
392 size_t rw_smallest_extent;
393 size_t rw_largest_extent;
394 struct ll_file_data *rw_last_file;
397 enum stats_track_type {
398 STATS_TRACK_ALL = 0, /* track all processes */
399 STATS_TRACK_PID, /* track process with this pid */
400 STATS_TRACK_PPID, /* track processes with this ppid */
401 STATS_TRACK_GID, /* track processes with this gid */
405 /* flags for sbi->ll_flags */
406 #define LL_SBI_NOLCK 0x01 /* DLM locking disabled (directio-only) */
407 #define LL_SBI_CHECKSUM 0x02 /* checksum each page as it's written */
408 #define LL_SBI_FLOCK 0x04
409 #define LL_SBI_USER_XATTR 0x08 /* support user xattr */
410 #define LL_SBI_ACL 0x10 /* support ACL */
411 #define LL_SBI_RMT_CLIENT 0x40 /* remote client */
412 #define LL_SBI_MDS_CAPA 0x80 /* support mds capa */
413 #define LL_SBI_OSS_CAPA 0x100 /* support oss capa */
414 #define LL_SBI_LOCALFLOCK 0x200 /* Local flocks support by kernel */
415 #define LL_SBI_LRU_RESIZE 0x400 /* lru resize support */
416 #define LL_SBI_LAZYSTATFS 0x800 /* lazystatfs mount option */
417 #define LL_SBI_SOM_PREVIEW 0x1000 /* SOM preview mount option */
418 #define LL_SBI_32BIT_API 0x2000 /* generate 32 bit inodes. */
419 #define LL_SBI_64BIT_HASH 0x4000 /* support 64-bits dir hash/offset */
420 #define LL_SBI_AGL_ENABLED 0x8000 /* enable agl */
421 #define LL_SBI_VERBOSE 0x10000 /* verbose mount/umount */
422 #define LL_SBI_LAYOUT_LOCK 0x20000 /* layout lock support */
423 #define LL_SBI_USER_FID2PATH 0x40000 /* allow fid2path by unprivileged users */
424 #define LL_SBI_XATTR_CACHE 0x80000 /* support for xattr cache */
425 #define LL_SBI_NOROOTSQUASH 0x100000 /* do not apply root squash */
427 #define LL_SBI_FLAGS { \
451 #define RCE_HASHES 32
453 struct rmtacl_ctl_entry {
454 struct list_head rce_list;
455 pid_t rce_key; /* hash key */
456 int rce_ops; /* acl operation type */
459 struct rmtacl_ctl_table {
461 struct list_head rct_entries[RCE_HASHES];
467 struct list_head ee_list;
468 pid_t ee_key; /* hash key */
469 struct lu_fid ee_fid;
470 int ee_type; /* ACL type for ACCESS or DEFAULT */
471 ext_acl_xattr_header *ee_acl;
476 struct list_head et_entries[EE_HASHES];
480 /* This is embedded into llite super-blocks to keep track of connect
481 * flags (capabilities) supported by all imports given mount is
483 struct lustre_client_ocd {
484 /* This is conjunction of connect_flags across all imports
485 * (LOVs) this mount is connected to. This field is updated by
486 * cl_ocd_update() under ->lco_lock. */
488 struct mutex lco_lock;
489 struct obd_export *lco_md_exp;
490 struct obd_export *lco_dt_exp;
494 struct list_head ll_list;
495 /* this protects pglist and ra_info. It isn't safe to
496 * grab from interrupt contexts */
498 spinlock_t ll_pp_extent_lock; /* pp_extent entry*/
499 spinlock_t ll_process_lock; /* ll_rw_process_info */
500 struct obd_uuid ll_sb_uuid;
501 struct obd_export *ll_md_exp;
502 struct obd_export *ll_dt_exp;
503 struct proc_dir_entry* ll_proc_root;
504 struct lu_fid ll_root_fid; /* root object fid */
507 unsigned int ll_umounting:1,
508 ll_xattr_cache_enabled:1;
509 /* per-conn chain of SBs */
510 struct list_head ll_conn_chain;
511 struct lustre_client_ocd ll_lco;
513 /*please don't ask -p*/
514 struct list_head ll_orphan_dentry_list;
515 struct ll_close_queue *ll_lcq;
517 struct lprocfs_stats *ll_stats; /* lprocfs stats counter */
519 /* Used to track "unstable" pages on a client, and maintain a
520 * LRU list of clean pages. An "unstable" page is defined as
521 * any page which is sent to a server as part of a bulk request,
522 * but is uncommitted to stable storage. */
523 struct cl_client_cache ll_cache;
525 struct lprocfs_stats *ll_ra_stats;
527 struct ll_ra_info ll_ra_info;
528 unsigned int ll_namelen;
529 struct file_operations *ll_fop;
531 /* =0 - hold lock over whole read/write
532 * >0 - max. chunk to be read/written w/o lock re-acquiring */
533 unsigned long ll_max_rw_chunk;
534 unsigned int ll_md_brw_pages; /* readdir pages per RPC */
536 struct lu_site *ll_site;
537 struct cl_device *ll_cl;
539 struct ll_rw_extents_info ll_rw_extents_info;
540 int ll_extent_process_count;
541 struct ll_rw_process_info ll_rw_process_info[LL_PROCESS_HIST_MAX];
542 unsigned int ll_offset_process_count;
543 struct ll_rw_process_info ll_rw_offset_info[LL_OFFSET_HIST_MAX];
544 unsigned int ll_rw_offset_entry_count;
545 int ll_stats_track_id;
546 enum stats_track_type ll_stats_track_type;
549 /* metadata stat-ahead */
550 unsigned int ll_sa_max; /* max statahead RPCs */
551 atomic_t ll_sa_total; /* statahead thread started
553 atomic_t ll_sa_wrong; /* statahead thread stopped for
555 atomic_t ll_sa_running; /* running statahead thread
557 atomic_t ll_agl_total; /* AGL thread started count */
559 dev_t ll_sdev_orig; /* save s_dev before assign for
561 struct rmtacl_ctl_table ll_rct;
562 struct eacl_table ll_et;
565 struct root_squash_info ll_squash;
568 #define LL_DEFAULT_MAX_RW_CHUNK (32 * 1024 * 1024)
571 * per file-descriptor read-ahead data.
573 struct ll_readahead_state {
576 * index of the last page that read(2) needed and that wasn't in the
577 * cache. Used by ras_update() to detect seeks.
579 * XXX nikita: if access seeks into cached region, Lustre doesn't see
582 unsigned long ras_last_readpage;
584 * number of pages read after last read-ahead window reset. As window
585 * is reset on each seek, this is effectively a number of consecutive
586 * accesses. Maybe ->ras_accessed_in_window is better name.
588 * XXX nikita: window is also reset (by ras_update()) when Lustre
589 * believes that memory pressure evicts read-ahead pages. In that
590 * case, it probably doesn't make sense to expand window to
591 * PTLRPC_MAX_BRW_PAGES on the third access.
593 unsigned long ras_consecutive_pages;
595 * number of read requests after the last read-ahead window reset
596 * As window is reset on each seek, this is effectively the number
597 * on consecutive read request and is used to trigger read-ahead.
599 unsigned long ras_consecutive_requests;
601 * Parameters of current read-ahead window. Handled by
602 * ras_update(). On the initial access to the file or after a seek,
603 * window is reset to 0. After 3 consecutive accesses, window is
604 * expanded to PTLRPC_MAX_BRW_PAGES. Afterwards, window is enlarged by
605 * PTLRPC_MAX_BRW_PAGES chunks up to ->ra_max_pages.
607 unsigned long ras_window_start, ras_window_len;
609 * Where next read-ahead should start at. This lies within read-ahead
610 * window. Read-ahead window is read in pieces rather than at once
611 * because: 1. lustre limits total number of pages under read-ahead by
612 * ->ra_max_pages (see ll_ra_count_get()), 2. client cannot read pages
613 * not covered by DLM lock.
615 unsigned long ras_next_readahead;
617 * Total number of ll_file_read requests issued, reads originating
618 * due to mmap are not counted in this total. This value is used to
619 * trigger full file read-ahead after multiple reads to a small file.
621 unsigned long ras_requests;
623 * Page index with respect to the current request, these value
624 * will not be accurate when dealing with reads issued via mmap.
626 unsigned long ras_request_index;
628 * The following 3 items are used for detecting the stride I/O
630 * In stride I/O mode,
631 * ...............|-----data-----|****gap*****|--------|******|....
632 * offset |-stride_pages-|-stride_gap-|
633 * ras_stride_offset = offset;
634 * ras_stride_length = stride_pages + stride_gap;
635 * ras_stride_pages = stride_pages;
636 * Note: all these three items are counted by pages.
638 unsigned long ras_stride_length;
639 unsigned long ras_stride_pages;
640 pgoff_t ras_stride_offset;
642 * number of consecutive stride request count, and it is similar as
643 * ras_consecutive_requests, but used for stride I/O mode.
644 * Note: only more than 2 consecutive stride request are detected,
645 * stride read-ahead will be enable
647 unsigned long ras_consecutive_stride_requests;
650 extern struct kmem_cache *ll_file_data_slab;
651 struct lustre_handle;
652 struct ll_file_data {
653 struct ll_readahead_state fd_ras;
654 struct ccc_grouplock fd_grouplock;
658 /* openhandle if lease exists for this file.
659 * Borrow lli->lli_och_mutex to protect assignment */
660 struct obd_client_handle *fd_lease_och;
661 struct obd_client_handle *fd_och;
662 struct file *fd_file;
663 /* Indicate whether need to report failure when close.
664 * true: failure is known, not report again.
665 * false: unknown failure, should report. */
666 bool fd_write_failed;
667 rwlock_t fd_lock; /* protect lcc list */
668 struct list_head fd_lccs; /* list of ll_cl_context */
671 struct lov_stripe_md;
673 extern spinlock_t inode_lock;
675 extern struct proc_dir_entry *proc_lustre_fs_root;
677 static inline struct inode *ll_info2i(struct ll_inode_info *lli)
679 return &lli->lli_vfs_inode;
682 __u32 ll_i2suppgid(struct inode *i);
683 void ll_i2gids(__u32 *suppgids, struct inode *i1,struct inode *i2);
685 static inline int ll_need_32bit_api(struct ll_sb_info *sbi)
687 #if BITS_PER_LONG == 32
689 #elif defined(CONFIG_COMPAT)
690 return unlikely(is_compat_task() || (sbi->ll_flags & LL_SBI_32BIT_API));
692 return unlikely(sbi->ll_flags & LL_SBI_32BIT_API);
696 void ll_ras_enter(struct file *f);
698 /* llite/lproc_llite.c */
700 int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
701 struct super_block *sb, char *osc, char *mdc);
702 void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi);
703 void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count);
704 extern struct lprocfs_seq_vars lprocfs_llite_obd_vars[];
706 static inline int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
707 struct super_block *sb, char *osc, char *mdc){return 0;}
708 static inline void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi) {}
709 static void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count) {}
714 LPROC_LL_DIRTY_MISSES,
716 LPROC_LL_WRITE_BYTES,
741 LPROC_LL_ALLOC_INODE,
744 LPROC_LL_GETXATTR_HITS,
746 LPROC_LL_REMOVEXATTR,
748 LPROC_LL_FILE_OPCODES
752 struct ll_dir_chain {
755 static inline void ll_dir_chain_init(struct ll_dir_chain *chain)
759 static inline void ll_dir_chain_fini(struct ll_dir_chain *chain)
763 extern const struct file_operations ll_dir_operations;
764 extern const struct inode_operations ll_dir_inode_operations;
765 #ifdef HAVE_DIR_CONTEXT
766 int ll_dir_read(struct inode *inode, __u64 *pos, struct md_op_data *op_data,
767 struct dir_context *ctx);
769 int ll_dir_read(struct inode *inode, __u64 *pos, struct md_op_data *op_data,
770 void *cookie, filldir_t filldir);
772 int ll_get_mdt_idx(struct inode *inode);
773 int ll_get_mdt_idx_by_fid(struct ll_sb_info *sbi, const struct lu_fid *fid);
774 struct page *ll_get_dir_page(struct inode *dir, struct md_op_data *op_data,
775 __u64 offset, struct ll_dir_chain *chain);
776 void ll_release_page(struct inode *inode, struct page *page, bool remove);
779 extern const struct inode_operations ll_special_inode_operations;
781 struct inode *ll_iget(struct super_block *sb, ino_t hash,
782 struct lustre_md *lic);
783 int ll_test_inode_by_fid(struct inode *inode, void *opaque);
784 int ll_md_blocking_ast(struct ldlm_lock *, struct ldlm_lock_desc *,
785 void *data, int flag);
786 struct dentry *ll_splice_alias(struct inode *inode, struct dentry *de);
787 int ll_rmdir_entry(struct inode *dir, char *name, int namelen);
788 void ll_update_times(struct ptlrpc_request *request, struct inode *inode);
791 int ll_writepage(struct page *page, struct writeback_control *wbc);
792 int ll_writepages(struct address_space *, struct writeback_control *wbc);
793 int ll_readpage(struct file *file, struct page *page);
794 void ll_readahead_init(struct inode *inode, struct ll_readahead_state *ras);
795 int ll_readahead(const struct lu_env *env, struct cl_io *io,
796 struct cl_page_list *queue, struct ll_readahead_state *ras,
798 int vvp_io_write_commit(const struct lu_env *env, struct cl_io *io);
799 struct ll_cl_context *ll_cl_find(struct file *file);
800 void ll_cl_add(struct file *file, const struct lu_env *env, struct cl_io *io);
801 void ll_cl_remove(struct file *file, const struct lu_env *env);
803 #ifndef MS_HAS_NEW_AOPS
804 extern const struct address_space_operations ll_aops;
806 extern const struct address_space_operations_ext ll_aops;
810 extern struct file_operations ll_file_operations;
811 extern struct file_operations ll_file_operations_flock;
812 extern struct file_operations ll_file_operations_noflock;
813 extern struct inode_operations ll_file_inode_operations;
814 extern int ll_have_md_lock(struct inode *inode, __u64 *bits,
815 ldlm_mode_t l_req_mode);
816 extern ldlm_mode_t ll_take_md_lock(struct inode *inode, __u64 bits,
817 struct lustre_handle *lockh, __u64 flags,
820 int ll_file_open(struct inode *inode, struct file *file);
821 int ll_file_release(struct inode *inode, struct file *file);
822 int ll_glimpse_ioctl(struct ll_sb_info *sbi,
823 struct lov_stripe_md *lsm, lstat_t *st);
824 void ll_ioepoch_open(struct ll_inode_info *lli, __u64 ioepoch);
825 int ll_release_openhandle(struct dentry *, struct lookup_intent *);
826 int ll_md_real_close(struct inode *inode, fmode_t fmode);
827 void ll_ioepoch_close(struct inode *inode, struct md_op_data *op_data,
828 struct obd_client_handle **och, unsigned long flags);
829 void ll_done_writing_attr(struct inode *inode, struct md_op_data *op_data);
830 int ll_som_update(struct inode *inode, struct md_op_data *op_data);
831 int ll_inode_getattr(struct inode *inode, struct obdo *obdo,
832 __u64 ioepoch, int sync);
833 void ll_pack_inode2opdata(struct inode *inode, struct md_op_data *op_data,
834 struct lustre_handle *fh);
835 extern void ll_rw_stats_tally(struct ll_sb_info *sbi, pid_t pid,
836 struct ll_file_data *file, loff_t pos,
837 size_t count, int rw);
838 int ll_getattr(struct vfsmount *mnt, struct dentry *de, struct kstat *stat);
839 struct posix_acl *ll_get_acl(struct inode *inode, int type);
840 int ll_migrate(struct inode *parent, struct file *file, int mdtidx,
841 const char *name, int namelen);
842 int ll_get_fid_by_name(struct inode *parent, const char *name,
843 int namelen, struct lu_fid *fid);
844 #ifdef HAVE_GENERIC_PERMISSION_4ARGS
845 int ll_inode_permission(struct inode *inode, int mask, unsigned int flags);
847 # ifndef HAVE_INODE_PERMISION_2ARGS
848 int ll_inode_permission(struct inode *inode, int mask, struct nameidata *nd);
850 int ll_inode_permission(struct inode *inode, int mask);
854 int ll_lov_setstripe_ea_info(struct inode *inode, struct file *file,
855 __u64 flags, struct lov_user_md *lum,
857 int ll_lov_getstripe_ea_info(struct inode *inode, const char *filename,
858 struct lov_mds_md **lmm, int *lmm_size,
859 struct ptlrpc_request **request);
860 int ll_dir_setstripe(struct inode *inode, struct lov_user_md *lump,
862 int ll_dir_getstripe(struct inode *inode, void **lmmp,
863 int *lmm_size, struct ptlrpc_request **request,
865 #ifdef HAVE_FILE_FSYNC_4ARGS
866 int ll_fsync(struct file *file, loff_t start, loff_t end, int data);
867 #elif defined(HAVE_FILE_FSYNC_2ARGS)
868 int ll_fsync(struct file *file, int data);
870 int ll_fsync(struct file *file, struct dentry *dentry, int data);
872 int ll_merge_attr(const struct lu_env *env, struct inode *inode);
873 int ll_fid2path(struct inode *inode, void __user *arg);
874 int ll_data_version(struct inode *inode, __u64 *data_version, int flags);
875 int ll_hsm_release(struct inode *inode);
879 int ll_d_init(struct dentry *de);
880 extern const struct dentry_operations ll_d_ops;
881 void ll_intent_drop_lock(struct lookup_intent *);
882 void ll_intent_release(struct lookup_intent *);
883 void ll_invalidate_aliases(struct inode *);
884 void ll_lookup_finish_locks(struct lookup_intent *it, struct dentry *dentry);
885 int ll_revalidate_it_finish(struct ptlrpc_request *request,
886 struct lookup_intent *it, struct dentry *de);
888 /* llite/llite_lib.c */
889 extern struct super_operations lustre_super_operations;
891 void ll_lli_init(struct ll_inode_info *lli);
892 int ll_fill_super(struct super_block *sb, struct vfsmount *mnt);
893 void ll_put_super(struct super_block *sb);
894 void ll_kill_super(struct super_block *sb);
895 struct inode *ll_inode_from_resource_lock(struct ldlm_lock *lock);
896 void ll_clear_inode(struct inode *inode);
897 int ll_setattr_raw(struct dentry *dentry, struct iattr *attr, bool hsm_import);
898 int ll_setattr(struct dentry *de, struct iattr *attr);
899 int ll_statfs(struct dentry *de, struct kstatfs *sfs);
900 int ll_statfs_internal(struct super_block *sb, struct obd_statfs *osfs,
901 __u64 max_age, __u32 flags);
902 int ll_update_inode(struct inode *inode, struct lustre_md *md);
903 int ll_read_inode2(struct inode *inode, void *opaque);
904 void ll_delete_inode(struct inode *inode);
905 int ll_iocontrol(struct inode *inode, struct file *file,
906 unsigned int cmd, unsigned long arg);
907 int ll_flush_ctx(struct inode *inode);
908 void ll_umount_begin(struct super_block *sb);
909 int ll_remount_fs(struct super_block *sb, int *flags, char *data);
910 #ifdef HAVE_SUPEROPS_USE_DENTRY
911 int ll_show_options(struct seq_file *seq, struct dentry *dentry);
913 int ll_show_options(struct seq_file *seq, struct vfsmount *vfs);
915 void ll_dirty_page_discard_warn(struct page *page, int ioret);
916 int ll_prep_inode(struct inode **inode, struct ptlrpc_request *req,
917 struct super_block *, struct lookup_intent *);
918 void lustre_dump_dentry(struct dentry *, int recur);
919 int ll_obd_statfs(struct inode *inode, void __user *arg);
920 int ll_get_max_mdsize(struct ll_sb_info *sbi, int *max_mdsize);
921 int ll_get_default_mdsize(struct ll_sb_info *sbi, int *default_mdsize);
922 int ll_get_max_cookiesize(struct ll_sb_info *sbi, int *max_cookiesize);
923 int ll_get_default_cookiesize(struct ll_sb_info *sbi, int *default_cookiesize);
924 int ll_process_config(struct lustre_cfg *lcfg);
925 struct md_op_data *ll_prep_md_op_data(struct md_op_data *op_data,
926 struct inode *i1, struct inode *i2,
927 const char *name, size_t namelen,
928 __u32 mode, __u32 opc, void *data);
929 void ll_finish_md_op_data(struct md_op_data *op_data);
930 int ll_get_obd_name(struct inode *inode, unsigned int cmd, unsigned long arg);
931 char *ll_get_fsname(struct super_block *sb, char *buf, int buflen);
932 void ll_compute_rootsquash_state(struct ll_sb_info *sbi);
933 ssize_t ll_copy_user_md(const struct lov_user_md __user *md,
934 struct lov_user_md **kbuf);
936 /* Compute expected user md size when passing in a md from user space */
937 static inline ssize_t ll_lov_user_md_size(const struct lov_user_md *lum)
939 switch (lum->lmm_magic) {
940 case LOV_USER_MAGIC_V1:
941 return sizeof(struct lov_user_md_v1);
942 case LOV_USER_MAGIC_V3:
943 return sizeof(struct lov_user_md_v3);
944 case LOV_USER_MAGIC_SPECIFIC:
945 if (lum->lmm_stripe_count > LOV_MAX_STRIPE_COUNT)
948 return lov_user_md_size(lum->lmm_stripe_count,
949 LOV_USER_MAGIC_SPECIFIC);
955 /* llite/llite_nfs.c */
956 extern struct export_operations lustre_export_operations;
957 __u32 get_uuid2int(const char *name, int len);
958 struct inode *search_inode_for_lustre(struct super_block *sb,
959 const struct lu_fid *fid);
960 int ll_dir_get_parent_fid(struct inode *dir, struct lu_fid *parent_fid);
962 /* llite/symlink.c */
963 extern struct inode_operations ll_fast_symlink_inode_operations;
965 /* llite/llite_close.c */
966 struct ll_close_queue {
968 struct list_head lcq_head;
969 wait_queue_head_t lcq_waitq;
970 struct completion lcq_comp;
974 void vvp_write_pending(struct vvp_object *club, struct vvp_page *page);
975 void vvp_write_complete(struct vvp_object *club, struct vvp_page *page);
977 /* specific achitecture can implement only part of this list */
978 enum vvp_io_subtype {
981 /** io started from splice_{read|write} */
988 enum vvp_io_subtype cui_io_subtype;
992 struct pipe_inode_info *cui_pipe;
993 unsigned int cui_flags;
995 struct vvp_fault_io {
997 * Inode modification time that is checked across DLM
1001 struct vm_area_struct *ft_vma;
1003 * locked page returned from vvp_io
1005 struct page *ft_vmpage;
1006 struct vm_fault_api {
1010 struct vm_fault *ft_vmf;
1012 * fault API used bitflags for return code.
1014 unsigned int ft_flags;
1016 * check that flags are from filemap_fault
1018 bool ft_flags_valid;
1023 /* Readahead state. */
1024 pgoff_t cui_ra_start;
1025 pgoff_t cui_ra_count;
1026 /* Set when cui_ra_{start,count} have been initialized. */
1031 * IO arguments for various VFS I/O interfaces.
1033 struct vvp_io_args {
1034 /** normal/sendfile/splice */
1035 enum vvp_io_subtype via_io_subtype;
1039 struct kiocb *via_iocb;
1040 struct iovec *via_iov;
1041 unsigned long via_nrsegs;
1044 struct pipe_inode_info *via_pipe;
1045 unsigned int via_flags;
1050 struct ll_cl_context {
1051 struct list_head lcc_list;
1053 const struct lu_env *lcc_env;
1054 struct cl_io *lcc_io;
1055 struct cl_page *lcc_page;
1058 struct vvp_thread_info {
1059 struct iovec vti_local_iov;
1060 struct vvp_io_args vti_args;
1061 struct ra_io_arg vti_ria;
1062 struct kiocb vti_kiocb;
1063 struct ll_cl_context vti_io_ctx;
1066 extern struct lu_context_key vvp_key;
1068 static inline struct vvp_thread_info *vvp_env_info(const struct lu_env *env)
1070 struct vvp_thread_info *info;
1072 info = lu_context_key_get(&env->le_ctx, &vvp_key);
1073 LASSERT(info != NULL);
1077 static inline struct vvp_io_args *vvp_env_args(const struct lu_env *env,
1078 enum vvp_io_subtype type)
1080 struct vvp_io_args *ret = &vvp_env_info(env)->vti_args;
1082 ret->via_io_subtype = type;
1087 struct vvp_session {
1088 struct vvp_io vs_ios;
1091 extern struct lu_context_key vvp_session_key;
1093 static inline struct vvp_session *vvp_env_session(const struct lu_env *env)
1095 struct vvp_session *ses;
1097 ses = lu_context_key_get(env->le_ses, &vvp_session_key);
1098 LASSERT(ses != NULL);
1102 static inline struct vvp_io *vvp_env_io(const struct lu_env *env)
1104 return &vvp_env_session(env)->vs_ios;
1107 int vvp_global_init(void);
1108 void vvp_global_fini(void);
1110 void ll_queue_done_writing(struct inode *inode, unsigned long flags);
1111 void ll_close_thread_shutdown(struct ll_close_queue *lcq);
1112 int ll_close_thread_start(struct ll_close_queue **lcq_ret);
1114 /* llite/llite_mmap.c */
1116 int ll_teardown_mmaps(struct address_space *mapping, __u64 first, __u64 last);
1117 int ll_file_mmap(struct file * file, struct vm_area_struct * vma);
1118 void policy_from_vma(ldlm_policy_data_t *policy,
1119 struct vm_area_struct *vma, unsigned long addr, size_t count);
1120 struct vm_area_struct *our_vma(struct mm_struct *mm, unsigned long addr,
1123 static inline void ll_invalidate_page(struct page *vmpage)
1125 struct address_space *mapping = vmpage->mapping;
1126 loff_t offset = vmpage->index << PAGE_CACHE_SHIFT;
1128 LASSERT(PageLocked(vmpage));
1129 if (mapping == NULL)
1133 * truncate_complete_page() calls
1134 * a_ops->invalidatepage()->cl_page_delete()->vvp_page_delete().
1136 ll_teardown_mmaps(mapping, offset, offset + PAGE_CACHE_SIZE);
1137 truncate_complete_page(mapping, vmpage);
1140 #define ll_s2sbi(sb) (s2lsi(sb)->lsi_llsbi)
1142 /* don't need an addref as the sb_info should be holding one */
1143 static inline struct obd_export *ll_s2dtexp(struct super_block *sb)
1145 return ll_s2sbi(sb)->ll_dt_exp;
1148 /* don't need an addref as the sb_info should be holding one */
1149 static inline struct obd_export *ll_s2mdexp(struct super_block *sb)
1151 return ll_s2sbi(sb)->ll_md_exp;
1154 static inline struct client_obd *sbi2mdc(struct ll_sb_info *sbi)
1156 struct obd_device *obd = sbi->ll_md_exp->exp_obd;
1162 // FIXME: replace the name of this with LL_SB to conform to kernel stuff
1163 static inline struct ll_sb_info *ll_i2sbi(struct inode *inode)
1165 return ll_s2sbi(inode->i_sb);
1168 static inline struct obd_export *ll_i2dtexp(struct inode *inode)
1170 return ll_s2dtexp(inode->i_sb);
1173 static inline struct obd_export *ll_i2mdexp(struct inode *inode)
1175 return ll_s2mdexp(inode->i_sb);
1178 static inline struct lu_fid *ll_inode2fid(struct inode *inode)
1182 LASSERT(inode != NULL);
1183 fid = &ll_i2info(inode)->lli_fid;
1188 static inline __u64 ll_file_maxbytes(struct inode *inode)
1190 return ll_i2info(inode)->lli_maxbytes;
1194 int ll_setxattr(struct dentry *dentry, const char *name,
1195 const void *value, size_t size, int flags);
1196 ssize_t ll_getxattr(struct dentry *dentry, const char *name,
1197 void *buffer, size_t size);
1198 ssize_t ll_listxattr(struct dentry *dentry, char *buffer, size_t size);
1199 int ll_removexattr(struct dentry *dentry, const char *name);
1201 /* llite/remote_perm.c */
1202 extern struct kmem_cache *ll_remote_perm_cachep;
1203 extern struct kmem_cache *ll_rmtperm_hash_cachep;
1205 void free_rmtperm_hash(struct hlist_head *hash);
1206 int ll_update_remote_perm(struct inode *inode, struct mdt_remote_perm *perm);
1207 int lustre_check_remote_perm(struct inode *inode, int mask);
1209 /* llite/llite_capa.c */
1210 extern struct timer_list ll_capa_timer;
1212 int ll_capa_thread_start(void);
1213 void ll_capa_thread_stop(void);
1214 void ll_capa_timer_callback(unsigned long unused);
1216 struct obd_capa *ll_add_capa(struct inode *inode, struct obd_capa *ocapa);
1218 void ll_capa_open(struct inode *inode);
1219 void ll_capa_close(struct inode *inode);
1221 struct obd_capa *ll_mdscapa_get(struct inode *inode);
1222 struct obd_capa *ll_osscapa_get(struct inode *inode, __u64 opc);
1224 void ll_truncate_free_capa(struct obd_capa *ocapa);
1225 void ll_clear_inode_capas(struct inode *inode);
1226 void ll_print_capa_stat(struct ll_sb_info *sbi);
1228 /* llite/llite_cl.c */
1229 extern struct lu_device_type vvp_device_type;
1232 * Common IO arguments for various VFS I/O interfaces.
1234 int cl_sb_init(struct super_block *sb);
1235 int cl_sb_fini(struct super_block *sb);
1237 void ras_update(struct ll_sb_info *sbi, struct inode *inode,
1238 struct ll_readahead_state *ras, unsigned long index,
1240 void ll_ra_count_put(struct ll_sb_info *sbi, unsigned long len);
1241 void ll_ra_stats_inc(struct inode *inode, enum ra_stat which);
1243 /* llite/llite_rmtacl.c */
1244 #ifdef CONFIG_FS_POSIX_ACL
1245 u64 rce_ops2valid(int ops);
1246 struct rmtacl_ctl_entry *rct_search(struct rmtacl_ctl_table *rct, pid_t key);
1247 int rct_add(struct rmtacl_ctl_table *rct, pid_t key, int ops);
1248 int rct_del(struct rmtacl_ctl_table *rct, pid_t key);
1249 void rct_init(struct rmtacl_ctl_table *rct);
1250 void rct_fini(struct rmtacl_ctl_table *rct);
1252 void ee_free(struct eacl_entry *ee);
1253 int ee_add(struct eacl_table *et, pid_t key, struct lu_fid *fid, int type,
1254 ext_acl_xattr_header *header);
1255 struct eacl_entry *et_search_del(struct eacl_table *et, pid_t key,
1256 struct lu_fid *fid, int type);
1257 void et_search_free(struct eacl_table *et, pid_t key);
1258 void et_init(struct eacl_table *et);
1259 void et_fini(struct eacl_table *et);
1264 #define LL_SA_RPC_MIN 2
1265 #define LL_SA_RPC_DEF 32
1266 #define LL_SA_RPC_MAX 8192
1268 #define LL_SA_CACHE_BIT 5
1269 #define LL_SA_CACHE_SIZE (1 << LL_SA_CACHE_BIT)
1270 #define LL_SA_CACHE_MASK (LL_SA_CACHE_SIZE - 1)
1272 /* per inode struct, for dir only */
1273 struct ll_statahead_info {
1274 struct dentry *sai_dentry;
1275 atomic_t sai_refcount; /* when access this struct, hold
1277 unsigned int sai_max; /* max ahead of lookup */
1278 __u64 sai_sent; /* stat requests sent count */
1279 __u64 sai_replied; /* stat requests which received
1281 __u64 sai_index; /* index of statahead entry */
1282 __u64 sai_index_wait; /* index of entry which is the
1283 * caller is waiting for */
1284 __u64 sai_hit; /* hit count */
1285 __u64 sai_miss; /* miss count:
1286 * for "ls -al" case, includes
1287 * hidden dentry miss;
1288 * for "ls -l" case, it does not
1289 * include hidden dentry miss.
1290 * "sai_miss_hidden" is used for
1293 unsigned int sai_consecutive_miss; /* consecutive miss */
1294 unsigned int sai_miss_hidden;/* "ls -al", but first dentry
1295 * is not a hidden one */
1296 unsigned int sai_skip_hidden;/* skipped hidden dentry count
1298 unsigned int sai_ls_all:1, /* "ls -al", do stat-ahead for
1300 sai_agl_valid:1,/* AGL is valid for the dir */
1301 sai_in_readpage:1;/* statahead is in readdir()*/
1302 wait_queue_head_t sai_waitq; /* stat-ahead wait queue */
1303 struct ptlrpc_thread sai_thread; /* stat-ahead thread */
1304 struct ptlrpc_thread sai_agl_thread; /* AGL thread */
1305 struct list_head sai_interim_entries; /* entries which got async
1306 * stat reply, but not
1308 struct list_head sai_entries; /* completed entries */
1309 struct list_head sai_agls; /* AGLs to be sent */
1310 struct list_head sai_cache[LL_SA_CACHE_SIZE];
1311 spinlock_t sai_cache_lock[LL_SA_CACHE_SIZE];
1312 atomic_t sai_cache_count; /* entry count in cache */
1315 int ll_statahead(struct inode *dir, struct dentry **dentry, bool unplug);
1316 void ll_authorize_statahead(struct inode *dir, void *key);
1317 void ll_deauthorize_statahead(struct inode *dir, void *key);
1319 static inline int ll_glimpse_size(struct inode *inode)
1321 struct ll_inode_info *lli = ll_i2info(inode);
1324 down_read(&lli->lli_glimpse_sem);
1325 rc = cl_glimpse_size(inode);
1326 lli->lli_glimpse_time = cfs_time_current();
1327 up_read(&lli->lli_glimpse_sem);
1331 /* dentry may statahead when statahead is enabled and current process has opened
1332 * parent directory, and this dentry hasn't accessed statahead cache before */
1334 dentry_may_statahead(struct inode *dir, struct dentry *dentry)
1336 struct ll_inode_info *lli;
1337 struct ll_dentry_data *ldd;
1339 if (ll_i2sbi(dir)->ll_sa_max == 0)
1342 lli = ll_i2info(dir);
1344 /* statahead is not allowed for this dir, there may be three causes:
1345 * 1. dir is not opened.
1346 * 2. statahead hit ratio is too low.
1347 * 3. previous stat started statahead thread failed. */
1348 if (!lli->lli_sa_enabled)
1351 /* not the same process, don't statahead */
1352 if (lli->lli_opendir_pid != current_pid())
1356 * When stating a dentry, kernel may trigger 'revalidate' or 'lookup'
1357 * multiple times, eg. for 'getattr', 'getxattr' and etc.
1358 * For patchless client, lookup intent is not accurate, which may
1359 * misguide statahead. For example:
1360 * The 'revalidate' call for 'getattr' and 'getxattr' of a dentry will
1361 * have the same intent -- IT_GETATTR, while one dentry should access
1362 * statahead cache once, otherwise statahead windows is messed up.
1363 * The solution is as following:
1364 * Assign 'lld_sa_generation' with 'lli_sa_generation' when a dentry
1365 * IT_GETATTR for the first time, and subsequent IT_GETATTR will
1366 * bypass interacting with statahead cache by checking
1367 * 'lld_sa_generation == lli->lli_sa_generation'.
1369 ldd = ll_d2d(dentry);
1370 if (ldd != NULL && ldd->lld_sa_generation == lli->lli_sa_generation)
1376 /* llite ioctl register support rountine */
1382 #define LLIOC_MAX_CMD 256
1385 * Rules to write a callback function:
1388 * @magic: Dynamic ioctl call routine will feed this vaule with the pointer
1389 * returned to ll_iocontrol_register. Callback functions should use this
1390 * data to check the potential collasion of ioctl cmd. If collasion is
1391 * found, callback function should return LLIOC_CONT.
1392 * @rcp: The result of ioctl command.
1395 * If @magic matches the pointer returned by ll_iocontrol_data, the
1396 * callback should return LLIOC_STOP; return LLIOC_STOP otherwise.
1398 typedef enum llioc_iter (*llioc_callback_t)(struct inode *inode,
1399 struct file *file, unsigned int cmd, unsigned long arg,
1400 void *magic, int *rcp);
1402 /* export functions */
1403 /* Register ioctl block dynamatically for a regular file.
1405 * @cmd: the array of ioctl command set
1406 * @count: number of commands in the @cmd
1407 * @cb: callback function, it will be called if an ioctl command is found to
1408 * belong to the command list @cmd.
1411 * A magic pointer will be returned if success;
1412 * otherwise, NULL will be returned.
1414 void *ll_iocontrol_register(llioc_callback_t cb, int count, unsigned int *cmd);
1415 void ll_iocontrol_unregister(void *magic);
1417 struct obd_capa *cl_capa_lookup(struct inode *inode, enum cl_req_type crt);
1419 int cl_sync_file_range(struct inode *inode, loff_t start, loff_t end,
1420 enum cl_fsync_mode mode, int ignore_layout);
1422 /** direct write pages */
1423 struct ll_dio_pages {
1424 /** page array to be written. we don't support
1425 * partial pages except the last one. */
1426 struct page **ldp_pages;
1427 /* offset of each page */
1428 loff_t *ldp_offsets;
1429 /** if ldp_offsets is NULL, it means a sequential
1430 * pages to be written, then this is the file offset
1431 * of the * first page. */
1432 loff_t ldp_start_offset;
1433 /** how many bytes are to be written. */
1435 /** # of pages in the array. */
1439 static inline void cl_stats_tally(struct cl_device *dev, enum cl_req_type crt,
1442 int opc = (crt == CRT_READ) ? LPROC_LL_OSC_READ :
1445 ll_stats_ops_tally(ll_s2sbi(cl2vvp_dev(dev)->vdv_sb), opc, rc);
1448 extern ssize_t ll_direct_rw_pages(const struct lu_env *env, struct cl_io *io,
1449 int rw, struct inode *inode,
1450 struct ll_dio_pages *pv);
1452 static inline int ll_file_nolock(const struct file *file)
1454 struct ll_file_data *fd = LUSTRE_FPRIVATE(file);
1455 struct inode *inode = file->f_dentry->d_inode;
1457 LASSERT(fd != NULL);
1458 return ((fd->fd_flags & LL_FILE_IGNORE_LOCK) ||
1459 (ll_i2sbi(inode)->ll_flags & LL_SBI_NOLCK));
1462 static inline void ll_set_lock_data(struct obd_export *exp, struct inode *inode,
1463 struct lookup_intent *it, __u64 *bits)
1465 if (!it->d.lustre.it_lock_set) {
1466 struct lustre_handle handle;
1468 /* If this inode is a remote object, it will get two
1469 * separate locks in different namespaces, Master MDT,
1470 * where the name entry is, will grant LOOKUP lock,
1471 * remote MDT, where the object is, will grant
1472 * UPDATE|PERM lock. The inode will be attched to both
1473 * LOOKUP and PERM locks, so revoking either locks will
1474 * case the dcache being cleared */
1475 if (it->d.lustre.it_remote_lock_mode) {
1476 handle.cookie = it->d.lustre.it_remote_lock_handle;
1477 CDEBUG(D_DLMTRACE, "setting l_data to inode "DFID
1478 "(%p) for remote lock "LPX64"\n",
1479 PFID(ll_inode2fid(inode)), inode,
1481 md_set_lock_data(exp, &handle.cookie, inode, NULL);
1484 handle.cookie = it->d.lustre.it_lock_handle;
1486 CDEBUG(D_DLMTRACE, "setting l_data to inode "DFID"(%p)"
1487 " for lock "LPX64"\n",
1488 PFID(ll_inode2fid(inode)), inode, handle.cookie);
1490 md_set_lock_data(exp, &handle.cookie, inode,
1491 &it->d.lustre.it_lock_bits);
1492 it->d.lustre.it_lock_set = 1;
1496 *bits = it->d.lustre.it_lock_bits;
1499 static inline void ll_lock_dcache(struct inode *inode)
1501 #ifdef HAVE_DCACHE_LOCK
1502 spin_lock(&dcache_lock);
1504 spin_lock(&inode->i_lock);
1508 static inline void ll_unlock_dcache(struct inode *inode)
1510 #ifdef HAVE_DCACHE_LOCK
1511 spin_unlock(&dcache_lock);
1513 spin_unlock(&inode->i_lock);
1517 static inline int d_lustre_invalid(const struct dentry *dentry)
1519 struct ll_dentry_data *lld = ll_d2d(dentry);
1521 return (lld == NULL) || lld->lld_invalid;
1524 static inline void __d_lustre_invalidate(struct dentry *dentry)
1526 struct ll_dentry_data *lld = ll_d2d(dentry);
1529 lld->lld_invalid = 1;
1533 * Mark dentry INVALID, if dentry refcount is zero (this is normally case for
1534 * ll_md_blocking_ast), unhash this dentry, and let dcache to reclaim it later;
1535 * else dput() of the last refcount will unhash this dentry and kill it.
1537 static inline void d_lustre_invalidate(struct dentry *dentry, int nested)
1539 CDEBUG(D_DENTRY, "invalidate dentry %.*s (%p) parent %p inode %p "
1540 "refc %d\n", dentry->d_name.len, dentry->d_name.name, dentry,
1541 dentry->d_parent, dentry->d_inode, ll_d_count(dentry));
1543 spin_lock_nested(&dentry->d_lock,
1544 nested ? DENTRY_D_LOCK_NESTED : DENTRY_D_LOCK_NORMAL);
1545 __d_lustre_invalidate(dentry);
1546 if (ll_d_count(dentry) == 0)
1548 spin_unlock(&dentry->d_lock);
1551 static inline void d_lustre_revalidate(struct dentry *dentry)
1553 spin_lock(&dentry->d_lock);
1554 LASSERT(ll_d2d(dentry) != NULL);
1555 ll_d2d(dentry)->lld_invalid = 0;
1556 spin_unlock(&dentry->d_lock);
1559 #if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 7, 53, 0)
1560 /* Compatibility for old (1.8) compiled userspace quota code */
1561 struct if_quotactl_18 {
1566 struct obd_dqinfo qc_dqinfo;
1567 struct obd_dqblk qc_dqblk;
1569 struct obd_uuid obd_uuid;
1571 #define LL_IOC_QUOTACTL_18 _IOWR('f', 162, struct if_quotactl_18 *)
1572 /* End compatibility for old (1.8) compiled userspace quota code */
1573 #endif /* LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 7, 53, 0) */
1576 LL_LAYOUT_GEN_NONE = ((__u32)-2), /* layout lock was cancelled */
1577 LL_LAYOUT_GEN_EMPTY = ((__u32)-1) /* for empty layout */
1580 int ll_layout_conf(struct inode *inode, const struct cl_object_conf *conf);
1581 int ll_layout_refresh(struct inode *inode, __u32 *gen);
1582 int ll_layout_restore(struct inode *inode, loff_t start, __u64 length);
1584 int ll_xattr_init(void);
1585 void ll_xattr_fini(void);
1587 int ll_page_sync_io(const struct lu_env *env, struct cl_io *io,
1588 struct cl_page *page, enum cl_req_type crt);
1590 int ll_getparent(struct file *file, struct getparent __user *arg);
1592 #endif /* LLITE_INTERNAL_H */