1 // -*- mode:C++; tab-width:8; c-basic-offset:2; indent-tabs-mode:t -*-
2 // vim: ts=8 sw=2 smarttab
4 * Ceph - scalable distributed file system
6 * Copyright (C) 2004-2006 Sage Weil <sage@newdream.net>
8 * This is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU Lesser General Public
10 * License version 2.1, as published by the Free Software
11 * Foundation. See file COPYING.
19 #include "common/CommandTable.h"
20 #include "common/Finisher.h"
21 #include "common/Timer.h"
22 #include "common/ceph_mutex.h"
23 #include "common/cmdparse.h"
24 #include "common/compiler_extensions.h"
25 #include "include/common_fwd.h"
26 #include "include/cephfs/ceph_ll_client.h"
27 #include "include/filepath.h"
28 #include "include/interval_set.h"
29 #include "include/lru.h"
30 #include "include/types.h"
31 #include "include/unordered_map.h"
32 #include "include/unordered_set.h"
33 #include "include/cephfs/metrics/Types.h"
34 #include "mds/mdstypes.h"
35 #include "msg/Dispatcher.h"
36 #include "msg/MessageRef.h"
37 #include "msg/Messenger.h"
38 #include "osdc/ObjectCacher.h"
42 #include "MetaSession.h"
67 class WritebackHandler
;
71 class destructive_lock_ref_t
;
84 class MDSCommandOp
: public CommandOp
89 explicit MDSCommandOp(ceph_tid_t t
) : CommandOp(t
) {}
92 /* error code for ceph_fuse */
93 #define CEPH_FUSE_NO_MDS_UP -((1<<16)+0) /* no mds up deteced in ceph_fuse */
94 #define CEPH_FUSE_LAST -((1<<16)+1) /* (unused) */
96 // ============================================
97 // types for my local metadata cache
100 - Dentries live in an LRU loop. they get expired based on last access.
101 see include/lru.h. items can be bumped to "mid" or "top" of list, etc.
102 - Inode has ref count for each Fh, Dir, or Dentry that points to it.
103 - when Inode ref goes to 0, it's expired.
104 - when Dir is empty, it's removed (and it's Inode ref--)
110 explicit DirEntry(const std::string
&s
) : d_name(s
), stmask(0) {}
111 DirEntry(const std::string
&n
, struct stat
& s
, int stm
)
112 : d_name(n
), st(s
), stmask(stm
) {}
127 class ceph_lock_state_t
;
129 // ========================================================
132 struct dir_result_t
{
133 static const int SHIFT
= 28;
134 static const int64_t MASK
= (1 << SHIFT
) - 1;
135 static const int64_t HASH
= 0xFFULL
<< (SHIFT
+ 24); // impossible frag bits
136 static const loff_t END
= 1ULL << (SHIFT
+ 32);
141 std::string alternate_name
;
143 explicit dentry(int64_t o
) : offset(o
) {}
144 dentry(int64_t o
, std::string n
, std::string an
, InodeRef in
) :
145 offset(o
), name(std::move(n
)), alternate_name(std::move(an
)), inode(std::move(in
)) {}
147 struct dentry_off_lt
{
148 bool operator()(const dentry
& d
, int64_t off
) const {
149 return dir_result_t::fpos_cmp(d
.offset
, off
) < 0;
154 explicit dir_result_t(Inode
*in
, const UserPerm
& perms
);
157 static uint64_t make_fpos(unsigned h
, unsigned l
, bool hash
) {
158 uint64_t v
= ((uint64_t)h
<< SHIFT
) | (uint64_t)l
;
162 ceph_assert((v
& HASH
) != HASH
);
165 static unsigned fpos_high(uint64_t p
) {
166 unsigned v
= (p
& (END
-1)) >> SHIFT
;
167 if ((p
& HASH
) == HASH
)
168 return ceph_frag_value(v
);
171 static unsigned fpos_low(uint64_t p
) {
174 static int fpos_cmp(uint64_t l
, uint64_t r
) {
175 int c
= ceph_frag_compare(fpos_high(l
), fpos_high(r
));
178 if (fpos_low(l
) == fpos_low(r
))
180 return fpos_low(l
) < fpos_low(r
) ? -1 : 1;
183 unsigned offset_high() { return fpos_high(offset
); }
184 unsigned offset_low() { return fpos_low(offset
); }
186 void set_end() { offset
|= END
; }
187 bool at_end() { return (offset
& END
); }
189 void set_hash_order() { offset
|= HASH
; }
190 bool hash_order() { return (offset
& HASH
) == HASH
; }
196 return buffer_frag
.contains(offset_high());
198 return buffer_frag
== frag_t(offset_high());
212 int64_t offset
; // hash order:
213 // (0xff << 52) | ((24 bits hash) << 28) |
214 // (the nth entry has hash collision);
216 // ((frag value) << 28) | (the nth entry in frag);
218 unsigned next_offset
; // offset of next chunk (last_name's + 1)
219 std::string last_name
; // last entry in previous chunk
221 uint64_t release_count
;
222 uint64_t ordered_count
;
223 unsigned cache_index
;
224 int start_shared_gen
; // dir shared_gen at start of readdir
229 std::vector
<dentry
> buffer
;
233 class Client
: public Dispatcher
, public md_config_obs_t
{
235 friend class C_Block_Sync
; // Calls block map and protected helpers
236 friend class C_Client_CacheInvalidate
; // calls ino_invalidate_cb
237 friend class C_Client_DentryInvalidate
; // calls dentry_invalidate_cb
238 friend class C_Client_FlushComplete
; // calls put_inode()
239 friend class C_Client_Remount
;
240 friend class C_Client_RequestInterrupt
;
241 friend class C_Deleg_Timeout
; // Asserts on client_lock, called when a delegation is unreturned
242 friend class C_Client_CacheRelease
; // Asserts on client_lock
243 friend class SyntheticClient
;
244 friend void intrusive_ptr_release(Inode
*in
);
245 template <typename T
> friend struct RWRefState
;
246 template <typename T
> friend class RWRef
;
248 using Dispatcher::cct
;
249 using clock
= ceph::coarse_mono_clock
;
251 typedef int (*add_dirent_cb_t
)(void *p
, struct dirent
*de
, struct ceph_statx
*stx
, off_t off
, Inode
*in
);
253 struct walk_dentry_result
{
255 std::string alternate_name
;
258 class CommandHook
: public AdminSocketHook
{
260 explicit CommandHook(Client
*client
);
261 int call(std::string_view command
, const cmdmap_t
& cmdmap
,
264 bufferlist
& out
) override
;
269 // snapshot info returned via get_snap_info(). nothing to do
270 // with SnapInfo on the MDS.
273 std::map
<std::string
, std::string
> metadata
;
276 Client(Messenger
*m
, MonClient
*mc
, Objecter
*objecter_
);
277 Client(const Client
&) = delete;
278 Client(const Client
&&) = delete;
279 virtual ~Client() override
;
281 static UserPerm
pick_my_perms(CephContext
*c
) {
282 uid_t uid
= c
->_conf
->client_mount_uid
>= 0 ? c
->_conf
->client_mount_uid
: -1;
283 gid_t gid
= c
->_conf
->client_mount_gid
>= 0 ? c
->_conf
->client_mount_gid
: -1;
284 return UserPerm(uid
, gid
);
286 UserPerm
pick_my_perms() {
287 uid_t uid
= user_id
>= 0 ? user_id
: -1;
288 gid_t gid
= group_id
>= 0 ? group_id
: -1;
289 return UserPerm(uid
, gid
);
292 int mount(const std::string
&mount_root
, const UserPerm
& perms
,
293 bool require_mds
=false, const std::string
&fs_name
="");
295 bool is_unmounting() const {
296 return mount_state
.check_current_state(CLIENT_UNMOUNTING
);
298 bool is_mounted() const {
299 return mount_state
.check_current_state(CLIENT_MOUNTED
);
301 bool is_mounting() const {
302 return mount_state
.check_current_state(CLIENT_MOUNTING
);
304 bool is_initialized() const {
305 return initialize_state
.check_current_state(CLIENT_INITIALIZED
);
309 void set_uuid(const std::string
& uuid
);
310 void set_session_timeout(unsigned timeout
);
311 int start_reclaim(const std::string
& uuid
, unsigned flags
,
312 const std::string
& fs_name
);
313 void finish_reclaim();
315 fs_cluster_id_t
get_fs_cid() {
320 const std::string
&mds_spec
,
321 const std::vector
<std::string
>& cmd
,
322 const bufferlist
& inbl
,
323 bufferlist
*poutbl
, std::string
*prs
, Context
*onfinish
);
325 // these should (more or less) mirror the actual system calls.
326 int statfs(const char *path
, struct statvfs
*stbuf
, const UserPerm
& perms
);
329 int chdir(const char *s
, std::string
&new_cwd
, const UserPerm
& perms
);
330 void _getcwd(std::string
& cwd
, const UserPerm
& perms
);
331 void getcwd(std::string
& cwd
, const UserPerm
& perms
);
334 int opendir(const char *name
, dir_result_t
**dirpp
, const UserPerm
& perms
);
335 int fdopendir(int dirfd
, dir_result_t
**dirpp
, const UserPerm
& perms
);
336 int closedir(dir_result_t
*dirp
);
339 * Fill a directory listing from dirp, invoking cb for each entry
340 * with the given pointer, the dirent, the struct stat, the stmask,
343 * Returns 0 if it reached the end of the directory.
344 * If @a cb returns a negative error code, stop and return that.
346 int readdir_r_cb(dir_result_t
*dirp
, add_dirent_cb_t cb
, void *p
,
347 unsigned want
=0, unsigned flags
=AT_NO_ATTR_SYNC
,
350 struct dirent
* readdir(dir_result_t
*d
);
351 int readdir_r(dir_result_t
*dirp
, struct dirent
*de
);
352 int readdirplus_r(dir_result_t
*dirp
, struct dirent
*de
, struct ceph_statx
*stx
, unsigned want
, unsigned flags
, Inode
**out
);
354 int getdir(const char *relpath
, std::list
<std::string
>& names
,
355 const UserPerm
& perms
); // get the whole dir at once.
358 * Returns the length of the buffer that got filled in, or -errno.
359 * If it returns -CEPHFS_ERANGE you just need to increase the size of the
360 * buffer and try again.
362 int _getdents(dir_result_t
*dirp
, char *buf
, int buflen
, bool ful
); // get a bunch of dentries at once
363 int getdents(dir_result_t
*dirp
, char *buf
, int buflen
) {
364 return _getdents(dirp
, buf
, buflen
, true);
366 int getdnames(dir_result_t
*dirp
, char *buf
, int buflen
) {
367 return _getdents(dirp
, buf
, buflen
, false);
370 void rewinddir(dir_result_t
*dirp
);
371 loff_t
telldir(dir_result_t
*dirp
);
372 void seekdir(dir_result_t
*dirp
, loff_t offset
);
374 int may_delete(const char *relpath
, const UserPerm
& perms
);
375 int link(const char *existing
, const char *newname
, const UserPerm
& perm
, std::string alternate_name
="");
376 int unlink(const char *path
, const UserPerm
& perm
);
377 int unlinkat(int dirfd
, const char *relpath
, int flags
, const UserPerm
& perm
);
378 int rename(const char *from
, const char *to
, const UserPerm
& perm
, std::string alternate_name
="");
381 int mkdir(const char *path
, mode_t mode
, const UserPerm
& perm
, std::string alternate_name
="");
382 int mkdirat(int dirfd
, const char *relpath
, mode_t mode
, const UserPerm
& perm
,
383 std::string alternate_name
="");
384 int mkdirs(const char *path
, mode_t mode
, const UserPerm
& perms
);
385 int rmdir(const char *path
, const UserPerm
& perms
);
388 int readlink(const char *path
, char *buf
, loff_t size
, const UserPerm
& perms
);
389 int readlinkat(int dirfd
, const char *relpath
, char *buf
, loff_t size
, const UserPerm
& perms
);
391 int symlink(const char *existing
, const char *newname
, const UserPerm
& perms
, std::string alternate_name
="");
392 int symlinkat(const char *target
, int dirfd
, const char *relpath
, const UserPerm
& perms
,
393 std::string alternate_name
="");
395 // path traversal for high-level interface
396 int walk(std::string_view path
, struct walk_dentry_result
* result
, const UserPerm
& perms
, bool followsym
=true);
399 unsigned statx_to_mask(unsigned int flags
, unsigned int want
);
400 int stat(const char *path
, struct stat
*stbuf
, const UserPerm
& perms
,
401 frag_info_t
*dirstat
=0, int mask
=CEPH_STAT_CAP_INODE_ALL
);
402 int statx(const char *path
, struct ceph_statx
*stx
,
403 const UserPerm
& perms
,
404 unsigned int want
, unsigned int flags
);
405 int lstat(const char *path
, struct stat
*stbuf
, const UserPerm
& perms
,
406 frag_info_t
*dirstat
=0, int mask
=CEPH_STAT_CAP_INODE_ALL
);
408 int setattr(const char *relpath
, struct stat
*attr
, int mask
,
409 const UserPerm
& perms
);
410 int setattrx(const char *relpath
, struct ceph_statx
*stx
, int mask
,
411 const UserPerm
& perms
, int flags
=0);
412 int fsetattr(int fd
, struct stat
*attr
, int mask
, const UserPerm
& perms
);
413 int fsetattrx(int fd
, struct ceph_statx
*stx
, int mask
, const UserPerm
& perms
);
414 int chmod(const char *path
, mode_t mode
, const UserPerm
& perms
);
415 int fchmod(int fd
, mode_t mode
, const UserPerm
& perms
);
416 int chmodat(int dirfd
, const char *relpath
, mode_t mode
, int flags
, const UserPerm
& perms
);
417 int lchmod(const char *path
, mode_t mode
, const UserPerm
& perms
);
418 int chown(const char *path
, uid_t new_uid
, gid_t new_gid
,
419 const UserPerm
& perms
);
420 int fchown(int fd
, uid_t new_uid
, gid_t new_gid
, const UserPerm
& perms
);
421 int lchown(const char *path
, uid_t new_uid
, gid_t new_gid
,
422 const UserPerm
& perms
);
423 int chownat(int dirfd
, const char *relpath
, uid_t new_uid
, gid_t new_gid
,
424 int flags
, const UserPerm
& perms
);
425 int utime(const char *path
, struct utimbuf
*buf
, const UserPerm
& perms
);
426 int lutime(const char *path
, struct utimbuf
*buf
, const UserPerm
& perms
);
427 int futime(int fd
, struct utimbuf
*buf
, const UserPerm
& perms
);
428 int utimes(const char *relpath
, struct timeval times
[2], const UserPerm
& perms
);
429 int lutimes(const char *relpath
, struct timeval times
[2], const UserPerm
& perms
);
430 int futimes(int fd
, struct timeval times
[2], const UserPerm
& perms
);
431 int futimens(int fd
, struct timespec times
[2], const UserPerm
& perms
);
432 int utimensat(int dirfd
, const char *relpath
, struct timespec times
[2], int flags
,
433 const UserPerm
& perms
);
434 int flock(int fd
, int operation
, uint64_t owner
);
435 int truncate(const char *path
, loff_t size
, const UserPerm
& perms
);
438 int mknod(const char *path
, mode_t mode
, const UserPerm
& perms
, dev_t rdev
=0);
440 int create_and_open(int dirfd
, const char *relpath
, int flags
, const UserPerm
& perms
,
441 mode_t mode
, int stripe_unit
, int stripe_count
, int object_size
,
442 const char *data_pool
, std::string alternate_name
);
443 int open(const char *path
, int flags
, const UserPerm
& perms
, mode_t mode
=0, std::string alternate_name
="") {
444 return open(path
, flags
, perms
, mode
, 0, 0, 0, NULL
, alternate_name
);
446 int open(const char *path
, int flags
, const UserPerm
& perms
,
447 mode_t mode
, int stripe_unit
, int stripe_count
, int object_size
,
448 const char *data_pool
, std::string alternate_name
="");
449 int openat(int dirfd
, const char *relpath
, int flags
, const UserPerm
& perms
,
450 mode_t mode
, int stripe_unit
, int stripe_count
,
451 int object_size
, const char *data_pool
, std::string alternate_name
);
452 int openat(int dirfd
, const char *path
, int flags
, const UserPerm
& perms
, mode_t mode
=0,
453 std::string alternate_name
="") {
454 return openat(dirfd
, path
, flags
, perms
, mode
, 0, 0, 0, NULL
, alternate_name
);
457 int lookup_hash(inodeno_t ino
, inodeno_t dirino
, const char *name
,
458 const UserPerm
& perms
);
459 int lookup_ino(inodeno_t ino
, const UserPerm
& perms
, Inode
**inode
=NULL
);
460 int lookup_name(Inode
*in
, Inode
*parent
, const UserPerm
& perms
);
463 loff_t
lseek(int fd
, loff_t offset
, int whence
);
464 int read(int fd
, char *buf
, loff_t size
, loff_t offset
=-1);
465 int preadv(int fd
, const struct iovec
*iov
, int iovcnt
, loff_t offset
=-1);
466 int write(int fd
, const char *buf
, loff_t size
, loff_t offset
=-1);
467 int pwritev(int fd
, const struct iovec
*iov
, int iovcnt
, loff_t offset
=-1);
468 int fake_write_size(int fd
, loff_t size
);
469 int ftruncate(int fd
, loff_t size
, const UserPerm
& perms
);
470 int fsync(int fd
, bool syncdataonly
);
471 int fstat(int fd
, struct stat
*stbuf
, const UserPerm
& perms
,
472 int mask
=CEPH_STAT_CAP_INODE_ALL
);
473 int fstatx(int fd
, struct ceph_statx
*stx
, const UserPerm
& perms
,
474 unsigned int want
, unsigned int flags
);
475 int statxat(int dirfd
, const char *relpath
,
476 struct ceph_statx
*stx
, const UserPerm
& perms
,
477 unsigned int want
, unsigned int flags
);
478 int fallocate(int fd
, int mode
, loff_t offset
, loff_t length
);
480 // full path xattr ops
481 int getxattr(const char *path
, const char *name
, void *value
, size_t size
,
482 const UserPerm
& perms
);
483 int lgetxattr(const char *path
, const char *name
, void *value
, size_t size
,
484 const UserPerm
& perms
);
485 int fgetxattr(int fd
, const char *name
, void *value
, size_t size
,
486 const UserPerm
& perms
);
487 int listxattr(const char *path
, char *list
, size_t size
, const UserPerm
& perms
);
488 int llistxattr(const char *path
, char *list
, size_t size
, const UserPerm
& perms
);
489 int flistxattr(int fd
, char *list
, size_t size
, const UserPerm
& perms
);
490 int removexattr(const char *path
, const char *name
, const UserPerm
& perms
);
491 int lremovexattr(const char *path
, const char *name
, const UserPerm
& perms
);
492 int fremovexattr(int fd
, const char *name
, const UserPerm
& perms
);
493 int setxattr(const char *path
, const char *name
, const void *value
,
494 size_t size
, int flags
, const UserPerm
& perms
);
495 int lsetxattr(const char *path
, const char *name
, const void *value
,
496 size_t size
, int flags
, const UserPerm
& perms
);
497 int fsetxattr(int fd
, const char *name
, const void *value
, size_t size
,
498 int flags
, const UserPerm
& perms
);
501 int64_t drop_caches();
503 int get_snap_info(const char *path
, const UserPerm
&perms
, SnapInfo
*snap_info
);
506 int lazyio(int fd
, int enable
);
507 int lazyio_propagate(int fd
, loff_t offset
, size_t count
);
508 int lazyio_synchronize(int fd
, loff_t offset
, size_t count
);
510 // expose file layout
511 int describe_layout(const char *path
, file_layout_t
* layout
,
512 const UserPerm
& perms
);
513 int fdescribe_layout(int fd
, file_layout_t
* layout
);
514 int get_file_stripe_address(int fd
, loff_t offset
, std::vector
<entity_addr_t
>& address
);
515 int get_file_extent_osds(int fd
, loff_t off
, loff_t
*len
, std::vector
<int>& osds
);
516 int get_osd_addr(int osd
, entity_addr_t
& addr
);
519 int64_t get_default_pool_id();
523 int get_pool_replication(int64_t pool
);
524 int64_t get_pool_id(const char *pool_name
);
525 std::string
get_pool_name(int64_t pool
);
526 int get_osd_crush_location(int id
, std::vector
<std::pair
<std::string
, std::string
> >& path
);
528 int enumerate_layout(int fd
, std::vector
<ObjectExtent
>& result
,
529 loff_t length
, loff_t offset
);
531 int mksnap(const char *path
, const char *name
, const UserPerm
& perm
,
532 mode_t mode
=0, const std::map
<std::string
, std::string
> &metadata
={});
533 int rmsnap(const char *path
, const char *name
, const UserPerm
& perm
, bool check_perms
=false);
535 // Inode permission checking
536 int inode_permission(Inode
*in
, const UserPerm
& perms
, unsigned want
);
539 int get_caps_issued(int fd
);
540 int get_caps_issued(const char *path
, const UserPerm
& perms
);
542 snapid_t
ll_get_snapid(Inode
*in
);
543 vinodeno_t
ll_get_vino(Inode
*in
) {
544 std::lock_guard
lock(client_lock
);
545 return _get_vino(in
);
547 // get inode from faked ino
548 Inode
*ll_get_inode(ino_t ino
);
549 Inode
*ll_get_inode(vinodeno_t vino
);
550 int ll_lookup(Inode
*parent
, const char *name
, struct stat
*attr
,
551 Inode
**out
, const UserPerm
& perms
);
552 int ll_lookup_inode(struct inodeno_t ino
, const UserPerm
& perms
, Inode
**inode
);
553 int ll_lookup_vino(vinodeno_t vino
, const UserPerm
& perms
, Inode
**inode
);
554 int ll_lookupx(Inode
*parent
, const char *name
, Inode
**out
,
555 struct ceph_statx
*stx
, unsigned want
, unsigned flags
,
556 const UserPerm
& perms
);
557 bool ll_forget(Inode
*in
, uint64_t count
);
558 bool ll_put(Inode
*in
);
559 int ll_get_snap_ref(snapid_t snap
);
561 int ll_getattr(Inode
*in
, struct stat
*st
, const UserPerm
& perms
);
562 int ll_getattrx(Inode
*in
, struct ceph_statx
*stx
, unsigned int want
,
563 unsigned int flags
, const UserPerm
& perms
);
564 int ll_setattrx(Inode
*in
, struct ceph_statx
*stx
, int mask
,
565 const UserPerm
& perms
);
566 int ll_setattr(Inode
*in
, struct stat
*st
, int mask
,
567 const UserPerm
& perms
);
568 int ll_getxattr(Inode
*in
, const char *name
, void *value
, size_t size
,
569 const UserPerm
& perms
);
570 int ll_setxattr(Inode
*in
, const char *name
, const void *value
, size_t size
,
571 int flags
, const UserPerm
& perms
);
572 int ll_removexattr(Inode
*in
, const char *name
, const UserPerm
& perms
);
573 int ll_listxattr(Inode
*in
, char *list
, size_t size
, const UserPerm
& perms
);
574 int ll_opendir(Inode
*in
, int flags
, dir_result_t
**dirpp
,
575 const UserPerm
& perms
);
576 int ll_releasedir(dir_result_t
* dirp
);
577 int ll_fsyncdir(dir_result_t
* dirp
);
578 int ll_readlink(Inode
*in
, char *buf
, size_t bufsize
, const UserPerm
& perms
);
579 int ll_mknod(Inode
*in
, const char *name
, mode_t mode
, dev_t rdev
,
580 struct stat
*attr
, Inode
**out
, const UserPerm
& perms
);
581 int ll_mknodx(Inode
*parent
, const char *name
, mode_t mode
, dev_t rdev
,
582 Inode
**out
, struct ceph_statx
*stx
, unsigned want
,
583 unsigned flags
, const UserPerm
& perms
);
584 int ll_mkdir(Inode
*in
, const char *name
, mode_t mode
, struct stat
*attr
,
585 Inode
**out
, const UserPerm
& perm
);
586 int ll_mkdirx(Inode
*parent
, const char *name
, mode_t mode
, Inode
**out
,
587 struct ceph_statx
*stx
, unsigned want
, unsigned flags
,
588 const UserPerm
& perms
);
589 int ll_symlink(Inode
*in
, const char *name
, const char *value
,
590 struct stat
*attr
, Inode
**out
, const UserPerm
& perms
);
591 int ll_symlinkx(Inode
*parent
, const char *name
, const char *value
,
592 Inode
**out
, struct ceph_statx
*stx
, unsigned want
,
593 unsigned flags
, const UserPerm
& perms
);
594 int ll_unlink(Inode
*in
, const char *name
, const UserPerm
& perm
);
595 int ll_rmdir(Inode
*in
, const char *name
, const UserPerm
& perms
);
596 int ll_rename(Inode
*parent
, const char *name
, Inode
*newparent
,
597 const char *newname
, const UserPerm
& perm
);
598 int ll_link(Inode
*in
, Inode
*newparent
, const char *newname
,
599 const UserPerm
& perm
);
600 int ll_open(Inode
*in
, int flags
, Fh
**fh
, const UserPerm
& perms
);
601 int _ll_create(Inode
*parent
, const char *name
, mode_t mode
,
602 int flags
, InodeRef
*in
, int caps
, Fh
**fhp
,
603 const UserPerm
& perms
);
604 int ll_create(Inode
*parent
, const char *name
, mode_t mode
, int flags
,
605 struct stat
*attr
, Inode
**out
, Fh
**fhp
,
606 const UserPerm
& perms
);
607 int ll_createx(Inode
*parent
, const char *name
, mode_t mode
,
608 int oflags
, Inode
**outp
, Fh
**fhp
,
609 struct ceph_statx
*stx
, unsigned want
, unsigned lflags
,
610 const UserPerm
& perms
);
611 int ll_read_block(Inode
*in
, uint64_t blockid
, char *buf
, uint64_t offset
,
612 uint64_t length
, file_layout_t
* layout
);
614 int ll_write_block(Inode
*in
, uint64_t blockid
,
615 char* buf
, uint64_t offset
,
616 uint64_t length
, file_layout_t
* layout
,
617 uint64_t snapseq
, uint32_t sync
);
618 int ll_commit_blocks(Inode
*in
, uint64_t offset
, uint64_t length
);
620 int ll_statfs(Inode
*in
, struct statvfs
*stbuf
, const UserPerm
& perms
);
621 int ll_walk(const char* name
, Inode
**i
, struct ceph_statx
*stx
,
622 unsigned int want
, unsigned int flags
, const UserPerm
& perms
);
623 uint32_t ll_stripe_unit(Inode
*in
);
624 int ll_file_layout(Inode
*in
, file_layout_t
*layout
);
625 uint64_t ll_snap_seq(Inode
*in
);
627 int ll_read(Fh
*fh
, loff_t off
, loff_t len
, bufferlist
*bl
);
628 int ll_write(Fh
*fh
, loff_t off
, loff_t len
, const char *data
);
629 int64_t ll_readv(struct Fh
*fh
, const struct iovec
*iov
, int iovcnt
, int64_t off
);
630 int64_t ll_writev(struct Fh
*fh
, const struct iovec
*iov
, int iovcnt
, int64_t off
);
631 loff_t
ll_lseek(Fh
*fh
, loff_t offset
, int whence
);
632 int ll_flush(Fh
*fh
);
633 int ll_fsync(Fh
*fh
, bool syncdataonly
);
634 int ll_sync_inode(Inode
*in
, bool syncdataonly
);
635 int ll_fallocate(Fh
*fh
, int mode
, int64_t offset
, int64_t length
);
636 int ll_release(Fh
*fh
);
637 int ll_getlk(Fh
*fh
, struct flock
*fl
, uint64_t owner
);
638 int ll_setlk(Fh
*fh
, struct flock
*fl
, uint64_t owner
, int sleep
);
639 int ll_flock(Fh
*fh
, int cmd
, uint64_t owner
);
640 int ll_lazyio(Fh
*fh
, int enable
);
641 int ll_file_layout(Fh
*fh
, file_layout_t
*layout
);
642 void ll_interrupt(void *d
);
643 bool ll_handle_umask() {
644 return acl_type
!= NO_ACL
;
647 int ll_get_stripe_osd(struct Inode
*in
, uint64_t blockno
,
648 file_layout_t
* layout
);
649 uint64_t ll_get_internal_offset(struct Inode
*in
, uint64_t blockno
);
651 int ll_num_osds(void);
652 int ll_osdaddr(int osd
, uint32_t *addr
);
653 int ll_osdaddr(int osd
, char* buf
, size_t size
);
655 void _ll_register_callbacks(struct ceph_client_callback_args
*args
);
656 void ll_register_callbacks(struct ceph_client_callback_args
*args
); // deprecated
657 int ll_register_callbacks2(struct ceph_client_callback_args
*args
);
658 std::pair
<int, bool> test_dentry_handling(bool can_invalidate
);
660 const char** get_tracked_conf_keys() const override
;
661 void handle_conf_change(const ConfigProxy
& conf
,
662 const std::set
<std::string
> &changed
) override
;
663 uint32_t get_deleg_timeout() { return deleg_timeout
; }
664 int set_deleg_timeout(uint32_t timeout
);
665 int ll_delegation(Fh
*fh
, unsigned cmd
, ceph_deleg_cb_t cb
, void *priv
);
667 entity_name_t
get_myname() { return messenger
->get_myname(); }
668 void wait_on_list(std::list
<ceph::condition_variable
*>& ls
);
669 void signal_cond_list(std::list
<ceph::condition_variable
*>& ls
);
671 void set_filer_flags(int flags
);
672 void clear_filer_flags(int flags
);
674 void tear_down_cache();
676 void update_metadata(std::string
const &k
, std::string
const &v
);
678 client_t
get_nodeid() { return whoami
; }
680 inodeno_t
get_root_ino();
684 virtual void shutdown();
687 void cancel_commands(const MDSMap
& newmap
);
688 void handle_mds_map(const MConstRef
<MMDSMap
>& m
);
689 void handle_fs_map(const MConstRef
<MFSMap
>& m
);
690 void handle_fs_map_user(const MConstRef
<MFSMapUser
>& m
);
691 void handle_osd_map(const MConstRef
<MOSDMap
>& m
);
693 void handle_lease(const MConstRef
<MClientLease
>& m
);
696 int uninline_data(Inode
*in
, Context
*onfinish
);
699 void check_cap_issue(Inode
*in
, unsigned issued
);
700 void add_update_cap(Inode
*in
, MetaSession
*session
, uint64_t cap_id
,
701 unsigned issued
, unsigned wanted
, unsigned seq
, unsigned mseq
,
702 inodeno_t realm
, int flags
, const UserPerm
& perms
);
703 void remove_cap(Cap
*cap
, bool queue_release
);
704 void remove_all_caps(Inode
*in
);
705 void remove_session_caps(MetaSession
*session
, int err
);
706 int mark_caps_flushing(Inode
*in
, ceph_tid_t
*ptid
);
707 void adjust_session_flushing_caps(Inode
*in
, MetaSession
*old_s
, MetaSession
*new_s
);
708 void flush_caps_sync();
709 void kick_flushing_caps(Inode
*in
, MetaSession
*session
);
710 void kick_flushing_caps(MetaSession
*session
);
711 void early_kick_flushing_caps(MetaSession
*session
);
712 int get_caps(Fh
*fh
, int need
, int want
, int *have
, loff_t endoff
);
713 int get_caps_used(Inode
*in
);
715 void maybe_update_snaprealm(SnapRealm
*realm
, snapid_t snap_created
, snapid_t snap_highwater
,
716 std::vector
<snapid_t
>& snaps
);
718 void handle_quota(const MConstRef
<MClientQuota
>& m
);
719 void handle_snap(const MConstRef
<MClientSnap
>& m
);
720 void handle_caps(const MConstRef
<MClientCaps
>& m
);
721 void handle_cap_import(MetaSession
*session
, Inode
*in
, const MConstRef
<MClientCaps
>& m
);
722 void handle_cap_export(MetaSession
*session
, Inode
*in
, const MConstRef
<MClientCaps
>& m
);
723 void handle_cap_trunc(MetaSession
*session
, Inode
*in
, const MConstRef
<MClientCaps
>& m
);
724 void handle_cap_flush_ack(MetaSession
*session
, Inode
*in
, Cap
*cap
, const MConstRef
<MClientCaps
>& m
);
725 void handle_cap_flushsnap_ack(MetaSession
*session
, Inode
*in
, const MConstRef
<MClientCaps
>& m
);
726 void handle_cap_grant(MetaSession
*session
, Inode
*in
, Cap
*cap
, const MConstRef
<MClientCaps
>& m
);
727 void cap_delay_requeue(Inode
*in
);
729 void send_cap(Inode
*in
, MetaSession
*session
, Cap
*cap
, int flags
,
730 int used
, int want
, int retain
, int flush
,
731 ceph_tid_t flush_tid
);
733 void send_flush_snap(Inode
*in
, MetaSession
*session
, snapid_t follows
, CapSnap
& capsnap
);
735 void flush_snaps(Inode
*in
);
736 void get_cap_ref(Inode
*in
, int cap
);
737 void put_cap_ref(Inode
*in
, int cap
);
738 void wait_sync_caps(Inode
*in
, ceph_tid_t want
);
739 void wait_sync_caps(ceph_tid_t want
);
740 void queue_cap_snap(Inode
*in
, SnapContext
&old_snapc
);
741 void finish_cap_snap(Inode
*in
, CapSnap
&capsnap
, int used
);
743 void _schedule_invalidate_dentry_callback(Dentry
*dn
, bool del
);
744 void _async_dentry_invalidate(vinodeno_t dirino
, vinodeno_t ino
, std::string
& name
);
745 void _try_to_trim_inode(Inode
*in
, bool sched_inval
);
747 void _schedule_invalidate_callback(Inode
*in
, int64_t off
, int64_t len
);
748 void _invalidate_inode_cache(Inode
*in
);
749 void _invalidate_inode_cache(Inode
*in
, int64_t off
, int64_t len
);
750 void _async_invalidate(vinodeno_t ino
, int64_t off
, int64_t len
);
752 void _schedule_ino_release_callback(Inode
*in
);
753 void _async_inode_release(vinodeno_t ino
);
755 bool _release(Inode
*in
);
758 * Initiate a flush of the data associated with the given inode.
759 * If you specify a Context, you are responsible for holding an inode
760 * reference for the duration of the flush. If not, _flush() will
761 * take the reference for you.
762 * @param in The Inode whose data you wish to flush.
763 * @param c The Context you wish us to complete once the data is
764 * flushed. If already flushed, this will be called in-line.
766 * @returns true if the data was already flushed, false otherwise.
768 bool _flush(Inode
*in
, Context
*c
);
769 void _flush_range(Inode
*in
, int64_t off
, uint64_t size
);
770 void _flushed(Inode
*in
);
771 void flush_set_callback(ObjectCacher::ObjectSet
*oset
);
773 void close_release(Inode
*in
);
774 void close_safe(Inode
*in
);
776 void lock_fh_pos(Fh
*f
);
777 void unlock_fh_pos(Fh
*f
);
780 void update_dir_dist(Inode
*in
, DirStat
*st
, mds_rank_t from
);
782 void clear_dir_complete_and_ordered(Inode
*diri
, bool complete
);
783 void insert_readdir_results(MetaRequest
*request
, MetaSession
*session
, Inode
*diri
);
784 Inode
* insert_trace(MetaRequest
*request
, MetaSession
*session
);
785 void update_inode_file_size(Inode
*in
, int issued
, uint64_t size
,
786 uint64_t truncate_seq
, uint64_t truncate_size
);
787 void update_inode_file_time(Inode
*in
, int issued
, uint64_t time_warp_seq
,
788 utime_t ctime
, utime_t mtime
, utime_t atime
);
790 Inode
*add_update_inode(InodeStat
*st
, utime_t ttl
, MetaSession
*session
,
791 const UserPerm
& request_perms
);
792 Dentry
*insert_dentry_inode(Dir
*dir
, const std::string
& dname
, LeaseStat
*dlease
,
793 Inode
*in
, utime_t from
, MetaSession
*session
,
794 Dentry
*old_dentry
= NULL
);
795 void update_dentry_lease(Dentry
*dn
, LeaseStat
*dlease
, utime_t from
, MetaSession
*session
);
797 bool use_faked_inos() { return _use_faked_inos
; }
798 vinodeno_t
map_faked_ino(ino_t ino
);
800 //notify the mds to flush the mdlog
801 void flush_mdlog_sync(Inode
*in
);
802 void flush_mdlog_sync();
803 void flush_mdlog(MetaSession
*session
);
806 void renew_caps(MetaSession
*session
);
807 void flush_cap_releases();
808 void renew_and_flush_cap_releases();
810 void start_tick_thread();
812 void update_read_io_size(size_t size
) {
814 total_read_size
+= size
;
817 void update_write_io_size(size_t size
) {
819 total_write_size
+= size
;
822 void inc_dentry_nr() {
825 void dec_dentry_nr() {
834 std::tuple
<uint64_t, uint64_t, uint64_t> get_dlease_hit_rates() {
835 return std::make_tuple(dlease_hits
, dlease_misses
, dentry_nr
);
844 std::pair
<uint64_t, uint64_t> get_cap_hit_rates() {
845 return std::make_pair(cap_hits
, cap_misses
);
848 void inc_opened_files() {
851 void dec_opened_files() {
854 std::pair
<uint64_t, uint64_t> get_opened_files_rates() {
855 return std::make_pair(opened_files
, inode_map
.size());
858 void inc_pinned_icaps() {
861 void dec_pinned_icaps(uint64_t nr
=1) {
864 std::pair
<uint64_t, uint64_t> get_pinned_icaps_rates() {
865 return std::make_pair(pinned_icaps
, inode_map
.size());
868 void inc_opened_inodes() {
871 void dec_opened_inodes() {
874 std::pair
<uint64_t, uint64_t> get_opened_inodes_rates() {
875 return std::make_pair(opened_inodes
, inode_map
.size());
878 /* timer_lock for 'timer' */
879 ceph::mutex timer_lock
= ceph::make_mutex("Client::timer_lock");
883 std::thread upkeeper
;
884 ceph::condition_variable upkeep_cond
;
885 bool tick_thread_stopped
= false;
887 std::unique_ptr
<PerfCounters
> logger
;
888 std::unique_ptr
<MDSMap
> mdsmap
;
890 bool fuse_default_permissions
;
891 bool _collect_and_send_global_metrics
;
894 /* Flags for check_caps() */
895 static const unsigned CHECK_CAPS_NODELAY
= 0x1;
896 static const unsigned CHECK_CAPS_SYNCHRONOUS
= 0x2;
898 void check_caps(Inode
*in
, unsigned flags
);
900 void set_cap_epoch_barrier(epoch_t e
);
902 void handle_command_reply(const MConstRef
<MCommandReply
>& m
);
903 int fetch_fsmap(bool user
);
905 const std::string
&mds_spec
,
906 std::vector
<mds_gid_t
> *targets
);
908 void get_session_metadata(std::map
<std::string
, std::string
> *meta
) const;
909 bool have_open_session(mds_rank_t mds
);
910 void got_mds_push(MetaSession
*s
);
911 MetaSessionRef
_get_mds_session(mds_rank_t mds
, Connection
*con
); ///< return session for mds *and* con; null otherwise
912 MetaSessionRef
_get_or_open_mds_session(mds_rank_t mds
);
913 MetaSessionRef
_open_mds_session(mds_rank_t mds
);
914 void _close_mds_session(MetaSession
*s
);
915 void _closed_mds_session(MetaSession
*s
, int err
=0, bool rejected
=false);
916 bool _any_stale_sessions() const;
917 void _kick_stale_sessions();
918 void handle_client_session(const MConstRef
<MClientSession
>& m
);
919 void send_reconnect(MetaSession
*s
);
920 void resend_unsafe_requests(MetaSession
*s
);
921 void wait_unsafe_requests();
923 void dump_mds_requests(Formatter
*f
);
924 void dump_mds_sessions(Formatter
*f
, bool cap_dump
=false);
926 int make_request(MetaRequest
*req
, const UserPerm
& perms
,
927 InodeRef
*ptarget
= 0, bool *pcreated
= 0,
928 mds_rank_t use_mds
=-1, bufferlist
*pdirbl
=0);
929 void put_request(MetaRequest
*request
);
930 void unregister_request(MetaRequest
*request
);
932 int verify_reply_trace(int r
, MetaSession
*session
, MetaRequest
*request
,
933 const MConstRef
<MClientReply
>& reply
,
934 InodeRef
*ptarget
, bool *pcreated
,
935 const UserPerm
& perms
);
936 void encode_cap_releases(MetaRequest
*request
, mds_rank_t mds
);
937 int encode_inode_release(Inode
*in
, MetaRequest
*req
,
938 mds_rank_t mds
, int drop
,
939 int unless
,int force
=0);
940 void encode_dentry_release(Dentry
*dn
, MetaRequest
*req
,
941 mds_rank_t mds
, int drop
, int unless
);
942 mds_rank_t
choose_target_mds(MetaRequest
*req
, Inode
** phash_diri
=NULL
);
943 void connect_mds_targets(mds_rank_t mds
);
944 void send_request(MetaRequest
*request
, MetaSession
*session
,
945 bool drop_cap_releases
=false);
946 MRef
<MClientRequest
> build_client_request(MetaRequest
*request
);
947 void kick_requests(MetaSession
*session
);
948 void kick_requests_closed(MetaSession
*session
);
949 void handle_client_request_forward(const MConstRef
<MClientRequestForward
>& reply
);
950 void handle_client_reply(const MConstRef
<MClientReply
>& reply
);
951 bool is_dir_operation(MetaRequest
*request
);
953 int path_walk(const filepath
& fp
, struct walk_dentry_result
* result
, const UserPerm
& perms
, bool followsym
=true, int mask
=0,
954 InodeRef dirinode
=nullptr);
955 int path_walk(const filepath
& fp
, InodeRef
*end
, const UserPerm
& perms
,
956 bool followsym
=true, int mask
=0, InodeRef dirinode
=nullptr);
958 // fake inode number for 32-bits ino_t
959 void _assign_faked_ino(Inode
*in
);
960 void _assign_faked_root(Inode
*in
);
961 void _release_faked_ino(Inode
*in
);
962 void _reset_faked_inos();
963 vinodeno_t
_map_faked_ino(ino_t ino
);
965 // Optional extra metadata about me to send to the MDS
966 void populate_metadata(const std::string
&mount_root
);
968 SnapRealm
*get_snap_realm(inodeno_t r
);
969 SnapRealm
*get_snap_realm_maybe(inodeno_t r
);
970 void put_snap_realm(SnapRealm
*realm
);
971 bool adjust_realm_parent(SnapRealm
*realm
, inodeno_t parent
);
972 void update_snap_trace(const bufferlist
& bl
, SnapRealm
**realm_ret
, bool must_flush
=true);
973 void invalidate_snaprealm_and_children(SnapRealm
*realm
);
975 Inode
*open_snapdir(Inode
*diri
);
978 int fd
= free_fd_set
.range_start();
979 free_fd_set
.erase(fd
, 1);
982 void put_fd(int fd
) {
983 free_fd_set
.insert(fd
, 1);
987 * Resolve file descriptor, or return NULL.
989 Fh
*get_filehandle(int fd
) {
990 auto it
= fd_map
.find(fd
);
991 if (it
== fd_map
.end())
995 int get_fd_inode(int fd
, InodeRef
*in
);
998 void wake_up_session_caps(MetaSession
*s
, bool reconnect
);
1000 void wait_on_context_list(std::list
<Context
*>& ls
);
1001 void signal_context_list(std::list
<Context
*>& ls
);
1003 // -- metadata cache stuff
1005 // decrease inode ref. delete if dangling.
1006 void _put_inode(Inode
*in
, int n
);
1007 void delay_put_inodes(bool wakeup
=false);
1008 void put_inode(Inode
*in
, int n
=1);
1009 void close_dir(Dir
*dir
);
1011 int subscribe_mdsmap(const std::string
&fs_name
="");
1013 void _abort_mds_sessions(int err
);
1015 // same as unmount() but for when the client_lock is already held
1016 void _unmount(bool abort
);
1018 //int get_cache_size() { return lru.lru_get_size(); }
1021 * Don't call this with in==NULL, use get_or_create for that
1022 * leave dn set to default NULL unless you're trying to add
1023 * a new inode to a pre-created Dentry
1025 Dentry
* link(Dir
*dir
, const std::string
& name
, Inode
*in
, Dentry
*dn
);
1026 void unlink(Dentry
*dn
, bool keepdir
, bool keepdentry
);
1028 int fill_stat(Inode
*in
, struct stat
*st
, frag_info_t
*dirstat
=0, nest_info_t
*rstat
=0);
1029 int fill_stat(InodeRef
& in
, struct stat
*st
, frag_info_t
*dirstat
=0, nest_info_t
*rstat
=0) {
1030 return fill_stat(in
.get(), st
, dirstat
, rstat
);
1033 void fill_statx(Inode
*in
, unsigned int mask
, struct ceph_statx
*stx
);
1034 void fill_statx(InodeRef
& in
, unsigned int mask
, struct ceph_statx
*stx
) {
1035 return fill_statx(in
.get(), mask
, stx
);
1038 void touch_dn(Dentry
*dn
);
1041 void trim_cache(bool trim_kernel_dcache
=false);
1042 void trim_cache_for_reconnect(MetaSession
*s
);
1043 void trim_dentry(Dentry
*dn
);
1044 void trim_caps(MetaSession
*s
, uint64_t max
);
1045 void _invalidate_kernel_dcache();
1046 void _trim_negative_child_dentries(InodeRef
& in
);
1048 void dump_inode(Formatter
*f
, Inode
*in
, set
<Inode
*>& did
, bool disconnected
);
1049 void dump_cache(Formatter
*f
); // debug
1052 void force_session_readonly(MetaSession
*s
);
1054 void dump_status(Formatter
*f
); // debug
1056 bool ms_dispatch2(const MessageRef
& m
) override
;
1058 void ms_handle_connect(Connection
*con
) override
;
1059 bool ms_handle_reset(Connection
*con
) override
;
1060 void ms_handle_remote_reset(Connection
*con
) override
;
1061 bool ms_handle_refused(Connection
*con
) override
;
1065 Inode
* get_quota_root(Inode
*in
, const UserPerm
& perms
);
1066 bool check_quota_condition(Inode
*in
, const UserPerm
& perms
,
1067 std::function
<bool (const Inode
&)> test
);
1068 bool is_quota_files_exceeded(Inode
*in
, const UserPerm
& perms
);
1069 bool is_quota_bytes_exceeded(Inode
*in
, int64_t new_bytes
,
1070 const UserPerm
& perms
);
1071 bool is_quota_bytes_approaching(Inode
*in
, const UserPerm
& perms
);
1073 int check_pool_perm(Inode
*in
, int need
);
1075 void handle_client_reclaim_reply(const MConstRef
<MClientReclaimReply
>& reply
);
1078 * Call this when an OSDMap is seen with a full flag (global or per pool)
1081 * @param pool the pool ID affected, or -1 if all.
1083 void _handle_full_flag(int64_t pool
);
1085 void _close_sessions();
1090 * The basic housekeeping parts of init (perf counters, admin socket)
1091 * that is independent of how objecters/monclient/messengers are
1094 void _finish_init();
1096 // global client lock
1097 // - protects Client and buffer cache both!
1098 ceph::mutex client_lock
= ceph::make_mutex("Client::client_lock");
1100 std::map
<snapid_t
, int> ll_snap_ref
;
1102 InodeRef root
= nullptr;
1103 map
<Inode
*, InodeRef
> root_parents
;
1104 Inode
* root_ancestor
= nullptr;
1105 LRU lru
; // lru list of Dentry's in our local metadata cache.
1109 std::unique_ptr
<Filer
> filer
;
1110 std::unique_ptr
<ObjectCacher
> objectcacher
;
1111 std::unique_ptr
<WritebackHandler
> writeback_handler
;
1113 Messenger
*messenger
;
1114 MonClient
*monclient
;
1119 /* The state migration mechanism */
1121 /* For the initialize_state */
1122 CLIENT_NEW
, // The initial state for the initialize_state or after Client::shutdown()
1123 CLIENT_INITIALIZING
, // At the beginning of the Client::init()
1124 CLIENT_INITIALIZED
, // At the end of CLient::init()
1126 /* For the mount_state */
1127 CLIENT_UNMOUNTED
, // The initial state for the mount_state or after unmounted
1128 CLIENT_MOUNTING
, // At the beginning of Client::mount()
1129 CLIENT_MOUNTED
, // At the end of Client::mount()
1130 CLIENT_UNMOUNTING
, // At the beginning of the Client::_unmout()
1133 typedef enum _state state_t
;
1134 using RWRef_t
= RWRef
<state_t
>;
1136 struct mount_state_t
: public RWRefState
<state_t
> {
1138 bool is_valid_state(state_t state
) const override
{
1140 case Client::CLIENT_MOUNTING
:
1141 case Client::CLIENT_MOUNTED
:
1142 case Client::CLIENT_UNMOUNTING
:
1143 case Client::CLIENT_UNMOUNTED
:
1150 int check_reader_state(state_t require
) const override
{
1151 if (require
== Client::CLIENT_MOUNTING
&&
1152 (state
== Client::CLIENT_MOUNTING
|| state
== Client::CLIENT_MOUNTED
))
1158 /* The state migration check */
1159 int check_writer_state(state_t require
) const override
{
1160 if (require
== Client::CLIENT_MOUNTING
&&
1161 state
== Client::CLIENT_UNMOUNTED
)
1163 else if (require
== Client::CLIENT_MOUNTED
&&
1164 state
== Client::CLIENT_MOUNTING
)
1166 else if (require
== Client::CLIENT_UNMOUNTING
&&
1167 state
== Client::CLIENT_MOUNTED
)
1169 else if (require
== Client::CLIENT_UNMOUNTED
&&
1170 state
== Client::CLIENT_UNMOUNTING
)
1176 mount_state_t(state_t state
, const char *lockname
, uint64_t reader_cnt
=0)
1177 : RWRefState (state
, lockname
, reader_cnt
) {}
1181 struct initialize_state_t
: public RWRefState
<state_t
> {
1183 bool is_valid_state(state_t state
) const override
{
1185 case Client::CLIENT_NEW
:
1186 case Client::CLIENT_INITIALIZING
:
1187 case Client::CLIENT_INITIALIZED
:
1194 int check_reader_state(state_t require
) const override
{
1195 if (require
== Client::CLIENT_INITIALIZED
&&
1196 state
>= Client::CLIENT_INITIALIZED
)
1202 /* The state migration check */
1203 int check_writer_state(state_t require
) const override
{
1204 if (require
== Client::CLIENT_INITIALIZING
&&
1205 (state
== Client::CLIENT_NEW
))
1207 else if (require
== Client::CLIENT_INITIALIZED
&&
1208 (state
== Client::CLIENT_INITIALIZING
))
1210 else if (require
== Client::CLIENT_NEW
&&
1211 (state
== Client::CLIENT_INITIALIZED
))
1217 initialize_state_t(state_t state
, const char *lockname
, uint64_t reader_cnt
=0)
1218 : RWRefState (state
, lockname
, reader_cnt
) {}
1219 ~initialize_state_t() {}
1222 struct mount_state_t mount_state
;
1223 struct initialize_state_t initialize_state
;
1226 struct C_Readahead
: public Context
{
1227 C_Readahead(Client
*c
, Fh
*f
);
1228 ~C_Readahead() override
;
1229 void finish(int r
) override
;
1236 * These define virtual xattrs exposing the recursive directory
1237 * statistics and layout metadata.
1240 const std::string name
;
1241 size_t (Client::*getxattr_cb
)(Inode
*in
, char *val
, size_t size
);
1243 bool (Client::*exists_cb
)(Inode
*in
);
1258 std::unique_ptr
<CephContext
, std::function
<void(CephContext
*)>> cct_deleter
;
1260 /* Flags for VXattr */
1261 static const unsigned VXATTR_RSTAT
= 0x1;
1262 static const unsigned VXATTR_DIRSTAT
= 0x2;
1264 static const VXattr _dir_vxattrs
[];
1265 static const VXattr _file_vxattrs
[];
1266 static const VXattr _common_vxattrs
[];
1269 bool is_reserved_vino(vinodeno_t
&vino
);
1271 void fill_dirent(struct dirent
*de
, const char *name
, int type
, uint64_t ino
, loff_t next_off
);
1273 int _opendir(Inode
*in
, dir_result_t
**dirpp
, const UserPerm
& perms
);
1274 void _readdir_drop_dirp_buffer(dir_result_t
*dirp
);
1275 bool _readdir_have_frag(dir_result_t
*dirp
);
1276 void _readdir_next_frag(dir_result_t
*dirp
);
1277 void _readdir_rechoose_frag(dir_result_t
*dirp
);
1278 int _readdir_get_frag(dir_result_t
*dirp
);
1279 int _readdir_cache_cb(dir_result_t
*dirp
, add_dirent_cb_t cb
, void *p
, int caps
, bool getref
);
1280 void _closedir(dir_result_t
*dirp
);
1283 void _fragmap_remove_non_leaves(Inode
*in
);
1284 void _fragmap_remove_stopped_mds(Inode
*in
, mds_rank_t mds
);
1286 void _ll_get(Inode
*in
);
1287 int _ll_put(Inode
*in
, uint64_t num
);
1288 void _ll_drop_pins();
1290 Fh
*_create_fh(Inode
*in
, int flags
, int cmode
, const UserPerm
& perms
);
1291 int _release_fh(Fh
*fh
);
1292 void _put_fh(Fh
*fh
);
1294 std::pair
<int, bool> _do_remount(bool retry_on_error
);
1296 int _read_sync(Fh
*f
, uint64_t off
, uint64_t len
, bufferlist
*bl
, bool *checkeof
);
1297 int _read_async(Fh
*f
, uint64_t off
, uint64_t len
, bufferlist
*bl
);
1299 bool _dentry_valid(const Dentry
*dn
);
1301 // internal interface
1302 // call these with client_lock held!
1303 int _do_lookup(Inode
*dir
, const std::string
& name
, int mask
, InodeRef
*target
,
1304 const UserPerm
& perms
);
1306 int _lookup(Inode
*dir
, const std::string
& dname
, int mask
, InodeRef
*target
,
1307 const UserPerm
& perm
, std::string
* alternate_name
=nullptr);
1309 int _link(Inode
*in
, Inode
*dir
, const char *name
, const UserPerm
& perm
, std::string alternate_name
,
1311 int _unlink(Inode
*dir
, const char *name
, const UserPerm
& perm
);
1312 int _rename(Inode
*olddir
, const char *oname
, Inode
*ndir
, const char *nname
, const UserPerm
& perm
, std::string alternate_name
);
1313 int _mkdir(Inode
*dir
, const char *name
, mode_t mode
, const UserPerm
& perm
,
1314 InodeRef
*inp
= 0, const std::map
<std::string
, std::string
> &metadata
={},
1315 std::string alternate_name
="");
1316 int _rmdir(Inode
*dir
, const char *name
, const UserPerm
& perms
);
1317 int _symlink(Inode
*dir
, const char *name
, const char *target
,
1318 const UserPerm
& perms
, std::string alternate_name
, InodeRef
*inp
= 0);
1319 int _mknod(Inode
*dir
, const char *name
, mode_t mode
, dev_t rdev
,
1320 const UserPerm
& perms
, InodeRef
*inp
= 0);
1321 int _do_setattr(Inode
*in
, struct ceph_statx
*stx
, int mask
,
1322 const UserPerm
& perms
, InodeRef
*inp
);
1323 void stat_to_statx(struct stat
*st
, struct ceph_statx
*stx
);
1324 int __setattrx(Inode
*in
, struct ceph_statx
*stx
, int mask
,
1325 const UserPerm
& perms
, InodeRef
*inp
= 0);
1326 int _setattrx(InodeRef
&in
, struct ceph_statx
*stx
, int mask
,
1327 const UserPerm
& perms
);
1328 int _setattr(InodeRef
&in
, struct stat
*attr
, int mask
,
1329 const UserPerm
& perms
);
1330 int _ll_setattrx(Inode
*in
, struct ceph_statx
*stx
, int mask
,
1331 const UserPerm
& perms
, InodeRef
*inp
= 0);
1332 int _getattr(Inode
*in
, int mask
, const UserPerm
& perms
, bool force
=false);
1333 int _getattr(InodeRef
&in
, int mask
, const UserPerm
& perms
, bool force
=false) {
1334 return _getattr(in
.get(), mask
, perms
, force
);
1336 int _readlink(Inode
*in
, char *buf
, size_t size
);
1337 int _getxattr(Inode
*in
, const char *name
, void *value
, size_t len
,
1338 const UserPerm
& perms
);
1339 int _getxattr(InodeRef
&in
, const char *name
, void *value
, size_t len
,
1340 const UserPerm
& perms
);
1341 int _getvxattr(Inode
*in
, const UserPerm
& perms
, const char *attr_name
,
1342 ssize_t size
, void *value
, mds_rank_t rank
);
1343 int _listxattr(Inode
*in
, char *names
, size_t len
, const UserPerm
& perms
);
1344 int _do_setxattr(Inode
*in
, const char *name
, const void *value
, size_t len
,
1345 int flags
, const UserPerm
& perms
);
1346 int _setxattr(Inode
*in
, const char *name
, const void *value
, size_t len
,
1347 int flags
, const UserPerm
& perms
);
1348 int _setxattr(InodeRef
&in
, const char *name
, const void *value
, size_t len
,
1349 int flags
, const UserPerm
& perms
);
1350 int _setxattr_check_data_pool(std::string
& name
, std::string
& value
, const OSDMap
*osdmap
);
1351 void _setxattr_maybe_wait_for_osdmap(const char *name
, const void *value
, size_t len
);
1352 int _removexattr(Inode
*in
, const char *nm
, const UserPerm
& perms
);
1353 int _removexattr(InodeRef
&in
, const char *nm
, const UserPerm
& perms
);
1354 int _open(Inode
*in
, int flags
, mode_t mode
, Fh
**fhp
,
1355 const UserPerm
& perms
);
1356 int _renew_caps(Inode
*in
);
1357 int _create(Inode
*in
, const char *name
, int flags
, mode_t mode
, InodeRef
*inp
,
1358 Fh
**fhp
, int stripe_unit
, int stripe_count
, int object_size
,
1359 const char *data_pool
, bool *created
, const UserPerm
&perms
,
1360 std::string alternate_name
);
1362 loff_t
_lseek(Fh
*fh
, loff_t offset
, int whence
);
1363 int64_t _read(Fh
*fh
, int64_t offset
, uint64_t size
, bufferlist
*bl
);
1364 int64_t _write(Fh
*fh
, int64_t offset
, uint64_t size
, const char *buf
,
1365 const struct iovec
*iov
, int iovcnt
);
1366 int64_t _preadv_pwritev_locked(Fh
*fh
, const struct iovec
*iov
,
1367 unsigned iovcnt
, int64_t offset
,
1368 bool write
, bool clamp_to_int
);
1369 int _preadv_pwritev(int fd
, const struct iovec
*iov
, unsigned iovcnt
,
1370 int64_t offset
, bool write
);
1372 int _fsync(Fh
*fh
, bool syncdataonly
);
1373 int _fsync(Inode
*in
, bool syncdataonly
);
1375 int _fallocate(Fh
*fh
, int mode
, int64_t offset
, int64_t length
);
1376 int _getlk(Fh
*fh
, struct flock
*fl
, uint64_t owner
);
1377 int _setlk(Fh
*fh
, struct flock
*fl
, uint64_t owner
, int sleep
);
1378 int _flock(Fh
*fh
, int cmd
, uint64_t owner
);
1379 int _lazyio(Fh
*fh
, int enable
);
1381 int get_or_create(Inode
*dir
, const char* name
,
1382 Dentry
**pdn
, bool expect_null
=false);
1384 int xattr_permission(Inode
*in
, const char *name
, unsigned want
,
1385 const UserPerm
& perms
);
1386 int may_setattr(Inode
*in
, struct ceph_statx
*stx
, int mask
,
1387 const UserPerm
& perms
);
1388 int may_open(Inode
*in
, int flags
, const UserPerm
& perms
);
1389 int may_lookup(Inode
*dir
, const UserPerm
& perms
);
1390 int may_create(Inode
*dir
, const UserPerm
& perms
);
1391 int may_delete(Inode
*dir
, const char *name
, const UserPerm
& perms
);
1392 int may_hardlink(Inode
*in
, const UserPerm
& perms
);
1394 int _getattr_for_perm(Inode
*in
, const UserPerm
& perms
);
1396 vinodeno_t
_get_vino(Inode
*in
);
1398 bool _vxattrcb_quota_exists(Inode
*in
);
1399 size_t _vxattrcb_quota(Inode
*in
, char *val
, size_t size
);
1400 size_t _vxattrcb_quota_max_bytes(Inode
*in
, char *val
, size_t size
);
1401 size_t _vxattrcb_quota_max_files(Inode
*in
, char *val
, size_t size
);
1403 bool _vxattrcb_layout_exists(Inode
*in
);
1404 size_t _vxattrcb_layout(Inode
*in
, char *val
, size_t size
);
1405 size_t _vxattrcb_layout_stripe_unit(Inode
*in
, char *val
, size_t size
);
1406 size_t _vxattrcb_layout_stripe_count(Inode
*in
, char *val
, size_t size
);
1407 size_t _vxattrcb_layout_object_size(Inode
*in
, char *val
, size_t size
);
1408 size_t _vxattrcb_layout_pool(Inode
*in
, char *val
, size_t size
);
1409 size_t _vxattrcb_layout_pool_namespace(Inode
*in
, char *val
, size_t size
);
1410 size_t _vxattrcb_dir_entries(Inode
*in
, char *val
, size_t size
);
1411 size_t _vxattrcb_dir_files(Inode
*in
, char *val
, size_t size
);
1412 size_t _vxattrcb_dir_subdirs(Inode
*in
, char *val
, size_t size
);
1413 size_t _vxattrcb_dir_rentries(Inode
*in
, char *val
, size_t size
);
1414 size_t _vxattrcb_dir_rfiles(Inode
*in
, char *val
, size_t size
);
1415 size_t _vxattrcb_dir_rsubdirs(Inode
*in
, char *val
, size_t size
);
1416 size_t _vxattrcb_dir_rsnaps(Inode
*in
, char *val
, size_t size
);
1417 size_t _vxattrcb_dir_rbytes(Inode
*in
, char *val
, size_t size
);
1418 size_t _vxattrcb_dir_rctime(Inode
*in
, char *val
, size_t size
);
1420 bool _vxattrcb_dir_pin_exists(Inode
*in
);
1421 size_t _vxattrcb_dir_pin(Inode
*in
, char *val
, size_t size
);
1423 bool _vxattrcb_snap_btime_exists(Inode
*in
);
1424 size_t _vxattrcb_snap_btime(Inode
*in
, char *val
, size_t size
);
1426 size_t _vxattrcb_caps(Inode
*in
, char *val
, size_t size
);
1428 bool _vxattrcb_mirror_info_exists(Inode
*in
);
1429 size_t _vxattrcb_mirror_info(Inode
*in
, char *val
, size_t size
);
1431 size_t _vxattrcb_cluster_fsid(Inode
*in
, char *val
, size_t size
);
1432 size_t _vxattrcb_client_id(Inode
*in
, char *val
, size_t size
);
1434 static const VXattr
*_get_vxattrs(Inode
*in
);
1435 static const VXattr
*_match_vxattr(Inode
*in
, const char *name
);
1437 int _do_filelock(Inode
*in
, Fh
*fh
, int lock_type
, int op
, int sleep
,
1438 struct flock
*fl
, uint64_t owner
, bool removing
=false);
1439 int _interrupt_filelock(MetaRequest
*req
);
1440 void _encode_filelocks(Inode
*in
, bufferlist
& bl
);
1441 void _release_filelocks(Fh
*fh
);
1442 void _update_lock_state(struct flock
*fl
, uint64_t owner
, ceph_lock_state_t
*lock_state
);
1444 int _posix_acl_create(Inode
*dir
, mode_t
*mode
, bufferlist
& xattrs_bl
,
1445 const UserPerm
& perms
);
1446 int _posix_acl_chmod(Inode
*in
, mode_t mode
, const UserPerm
& perms
);
1447 int _posix_acl_permission(Inode
*in
, const UserPerm
& perms
, unsigned want
);
1449 mds_rank_t
_get_random_up_mds() const;
1451 int _ll_getattr(Inode
*in
, int caps
, const UserPerm
& perms
);
1452 int _lookup_parent(Inode
*in
, const UserPerm
& perms
, Inode
**parent
=NULL
);
1453 int _lookup_name(Inode
*in
, Inode
*parent
, const UserPerm
& perms
);
1454 int _lookup_vino(vinodeno_t ino
, const UserPerm
& perms
, Inode
**inode
=NULL
);
1455 bool _ll_forget(Inode
*in
, uint64_t count
);
1457 void collect_and_send_metrics();
1458 void collect_and_send_global_metrics();
1460 uint32_t deleg_timeout
= 0;
1462 client_switch_interrupt_callback_t switch_interrupt_cb
= nullptr;
1463 client_remount_callback_t remount_cb
= nullptr;
1464 client_ino_callback_t ino_invalidate_cb
= nullptr;
1465 client_dentry_callback_t dentry_invalidate_cb
= nullptr;
1466 client_umask_callback_t umask_cb
= nullptr;
1467 client_ino_release_t ino_release_cb
= nullptr;
1468 void *callback_handle
= nullptr;
1469 bool can_invalidate_dentries
= false;
1471 Finisher async_ino_invalidator
;
1472 Finisher async_dentry_invalidator
;
1473 Finisher interrupt_finisher
;
1474 Finisher remount_finisher
;
1475 Finisher async_ino_releasor
;
1476 Finisher objecter_finisher
;
1478 utime_t last_cap_renew
;
1480 CommandHook m_command_hook
;
1482 int user_id
, group_id
;
1483 int acl_type
= NO_ACL
;
1485 epoch_t cap_epoch_barrier
= 0;
1488 map
<mds_rank_t
, MetaSessionRef
> mds_sessions
; // mds -> push seq
1489 std::set
<mds_rank_t
> mds_ranks_closing
; // mds ranks currently tearing down sessions
1490 std::list
<ceph::condition_variable
*> waiting_for_mdsmap
;
1492 // FSMap, for when using mds_command
1493 std::list
<ceph::condition_variable
*> waiting_for_fsmap
;
1494 std::unique_ptr
<FSMap
> fsmap
;
1495 std::unique_ptr
<FSMapUser
> fsmap_user
;
1497 // This mutex only protects command_table
1498 ceph::mutex command_lock
= ceph::make_mutex("Client::command_lock");
1499 // MDS command state
1500 CommandTable
<MDSCommandOp
> command_table
;
1502 bool _use_faked_inos
;
1505 fs_cluster_id_t fscid
;
1507 // file handles, etc.
1508 interval_set
<int> free_fd_set
; // unused fds
1509 ceph::unordered_map
<int, Fh
*> fd_map
;
1510 set
<Fh
*> ll_unclosed_fh_set
;
1511 ceph::unordered_set
<dir_result_t
*> opened_dirs
;
1512 uint64_t fd_gen
= 1;
1514 bool mount_aborted
= false;
1515 bool blocklisted
= false;
1517 ceph::unordered_map
<vinodeno_t
, Inode
*> inode_map
;
1518 ceph::unordered_map
<ino_t
, vinodeno_t
> faked_ino_map
;
1519 interval_set
<ino_t
> free_faked_inos
;
1520 ino_t last_used_faked_ino
;
1521 ino_t last_used_faked_root
;
1523 int local_osd
= -CEPHFS_ENXIO
;
1524 epoch_t local_osd_epoch
= 0;
1527 ceph_tid_t last_tid
= 0;
1528 ceph_tid_t oldest_tid
= 0; // oldest incomplete mds request, excluding setfilelock requests
1529 map
<ceph_tid_t
, MetaRequest
*> mds_requests
;
1532 ceph_tid_t last_flush_tid
= 1;
1534 xlist
<Inode
*> delayed_list
;
1535 int num_flushing_caps
= 0;
1536 ceph::unordered_map
<inodeno_t
,SnapRealm
*> snap_realms
;
1537 std::map
<std::string
, std::string
> metadata
;
1539 utime_t last_auto_reconnect
;
1542 std::ofstream traceout
;
1544 ceph::condition_variable mount_cond
, sync_cond
;
1546 std::map
<std::pair
<int64_t,std::string
>, int> pool_perms
;
1547 std::list
<ceph::condition_variable
*> waiting_for_pool_perm
;
1549 uint64_t retries_on_invalidate
= 0;
1552 std::list
<ceph::condition_variable
*> waiting_for_reclaim
;
1553 int reclaim_errno
= 0;
1554 epoch_t reclaim_osd_epoch
= 0;
1555 entity_addrvec_t reclaim_target_addrs
;
1557 // dentry lease metrics
1558 uint64_t dentry_nr
= 0;
1559 uint64_t dlease_hits
= 0;
1560 uint64_t dlease_misses
= 0;
1562 uint64_t cap_hits
= 0;
1563 uint64_t cap_misses
= 0;
1565 uint64_t opened_files
= 0;
1566 uint64_t pinned_icaps
= 0;
1567 uint64_t opened_inodes
= 0;
1569 uint64_t total_read_ops
= 0;
1570 uint64_t total_read_size
= 0;
1572 uint64_t total_write_ops
= 0;
1573 uint64_t total_write_size
= 0;
1575 ceph::spinlock delay_i_lock
;
1576 std::map
<Inode
*,int> delay_i_release
;
1580 * Specialization of Client that manages its own Objecter instance
1581 * and handles init/shutdown of messenger/monclient
1583 class StandaloneClient
: public Client
1586 StandaloneClient(Messenger
*m
, MonClient
*mc
, boost::asio::io_context
& ictx
);
1588 ~StandaloneClient() override
;
1590 int init() override
;
1591 void shutdown() override
;