Commit | Line | Data |
---|---|---|
b2441318 | 1 | /* SPDX-License-Identifier: GPL-2.0 */ |
2f2dc053 SW |
2 | #ifndef _FS_CEPH_MDS_CLIENT_H |
3 | #define _FS_CEPH_MDS_CLIENT_H | |
4 | ||
5 | #include <linux/completion.h> | |
153c8e6b | 6 | #include <linux/kref.h> |
2f2dc053 SW |
7 | #include <linux/list.h> |
8 | #include <linux/mutex.h> | |
44ca18f2 | 9 | #include <linux/rbtree.h> |
2f2dc053 | 10 | #include <linux/spinlock.h> |
3997c01d | 11 | #include <linux/refcount.h> |
717e6f28 | 12 | #include <linux/utsname.h> |
70c94820 | 13 | #include <linux/ktime.h> |
2f2dc053 | 14 | |
3d14c5d2 YS |
15 | #include <linux/ceph/types.h> |
16 | #include <linux/ceph/messenger.h> | |
6c4a1915 | 17 | #include <linux/ceph/auth.h> |
2f2dc053 | 18 | |
522dc510 | 19 | #include "mdsmap.h" |
f9009efa | 20 | #include "metric.h" |
3e699bd8 | 21 | #include "super.h" |
f9009efa | 22 | |
342ce182 | 23 | /* The first 8 bits are reserved for old ceph releases */ |
9ba1e224 XL |
24 | enum ceph_feature_type { |
25 | CEPHFS_FEATURE_MIMIC = 8, | |
26 | CEPHFS_FEATURE_REPLY_ENCODING, | |
27 | CEPHFS_FEATURE_RECLAIM_CLIENT, | |
28 | CEPHFS_FEATURE_LAZY_CAP_WANTED, | |
29 | CEPHFS_FEATURE_MULTI_RECONNECT, | |
d4846487 | 30 | CEPHFS_FEATURE_DELEG_INO, |
18f473b3 | 31 | CEPHFS_FEATURE_METRIC_COLLECT, |
300e42a2 XL |
32 | CEPHFS_FEATURE_ALTERNATE_NAME, |
33 | CEPHFS_FEATURE_NOTIFY_SESSION_STATE, | |
6eb06c46 | 34 | CEPHFS_FEATURE_OP_GETVXATTR, |
ce0d5bd3 | 35 | CEPHFS_FEATURE_32BITS_RETRY_FWD, |
5ccd8530 CB |
36 | CEPHFS_FEATURE_NEW_SNAPREALM_INFO, |
37 | CEPHFS_FEATURE_HAS_OWNER_UIDGID, | |
d2f8bb27 | 38 | |
5ccd8530 | 39 | CEPHFS_FEATURE_MAX = CEPHFS_FEATURE_HAS_OWNER_UIDGID, |
9ba1e224 XL |
40 | }; |
41 | ||
9ba1e224 | 42 | #define CEPHFS_FEATURES_CLIENT_SUPPORTED { \ |
d2f8bb27 YZ |
43 | 0, 1, 2, 3, 4, 5, 6, 7, \ |
44 | CEPHFS_FEATURE_MIMIC, \ | |
b37fe1f9 | 45 | CEPHFS_FEATURE_REPLY_ENCODING, \ |
d2f8bb27 | 46 | CEPHFS_FEATURE_LAZY_CAP_WANTED, \ |
81c5a148 | 47 | CEPHFS_FEATURE_MULTI_RECONNECT, \ |
d4846487 | 48 | CEPHFS_FEATURE_DELEG_INO, \ |
18f473b3 | 49 | CEPHFS_FEATURE_METRIC_COLLECT, \ |
4ac4c23e | 50 | CEPHFS_FEATURE_ALTERNATE_NAME, \ |
300e42a2 | 51 | CEPHFS_FEATURE_NOTIFY_SESSION_STATE, \ |
6eb06c46 | 52 | CEPHFS_FEATURE_OP_GETVXATTR, \ |
ce0d5bd3 | 53 | CEPHFS_FEATURE_32BITS_RETRY_FWD, \ |
5ccd8530 | 54 | CEPHFS_FEATURE_HAS_OWNER_UIDGID, \ |
342ce182 | 55 | } |
342ce182 | 56 | |
2f2dc053 SW |
57 | /* |
58 | * Some lock dependencies: | |
59 | * | |
60 | * session->s_mutex | |
61 | * mdsc->mutex | |
62 | * | |
63 | * mdsc->snap_rwsem | |
64 | * | |
be655596 | 65 | * ci->i_ceph_lock |
2f2dc053 SW |
66 | * mdsc->snap_flush_lock |
67 | * mdsc->cap_delay_lock | |
68 | * | |
69 | */ | |
70 | ||
3d14c5d2 | 71 | struct ceph_fs_client; |
2f2dc053 SW |
72 | struct ceph_cap; |
73 | ||
74 | /* | |
75 | * parsed info about a single inode. pointers are into the encoded | |
76 | * on-wire structures within the mds reply message payload. | |
77 | */ | |
78 | struct ceph_mds_reply_info_in { | |
79 | struct ceph_mds_reply_inode *in; | |
14303d20 | 80 | struct ceph_dir_layout dir_layout; |
2f2dc053 SW |
81 | u32 symlink_len; |
82 | char *symlink; | |
83 | u32 xattr_len; | |
84 | char *xattr_data; | |
fb01d1f8 YZ |
85 | u64 inline_version; |
86 | u32 inline_len; | |
87 | char *inline_data; | |
5ea5c5e0 | 88 | u32 pool_ns_len; |
779fe0fb | 89 | char *pool_ns_data; |
fb18a575 LH |
90 | u64 max_bytes; |
91 | u64 max_files; | |
08796873 | 92 | s32 dir_pin; |
245ce991 | 93 | struct ceph_timespec btime; |
193e7b37 | 94 | struct ceph_timespec snap_btime; |
2d332d5b JL |
95 | u8 *fscrypt_auth; |
96 | u8 *fscrypt_file; | |
97 | u32 fscrypt_auth_len; | |
98 | u32 fscrypt_file_len; | |
e7f72952 | 99 | u64 rsnaps; |
a35ead31 | 100 | u64 change_attr; |
2f2dc053 SW |
101 | }; |
102 | ||
2a5beea3 | 103 | struct ceph_mds_reply_dir_entry { |
af9ffa6d | 104 | bool is_nokey; |
2a5beea3 YZ |
105 | char *name; |
106 | u32 name_len; | |
af9ffa6d | 107 | u32 raw_hash; |
2a5beea3 YZ |
108 | struct ceph_mds_reply_lease *lease; |
109 | struct ceph_mds_reply_info_in inode; | |
8974eebd | 110 | loff_t offset; |
2a5beea3 YZ |
111 | }; |
112 | ||
6ddf5f16 MC |
113 | struct ceph_mds_reply_xattr { |
114 | char *xattr_value; | |
115 | size_t xattr_value_len; | |
116 | }; | |
117 | ||
2f2dc053 | 118 | /* |
25933abd HS |
119 | * parsed info about an mds reply, including information about |
120 | * either: 1) the target inode and/or its parent directory and dentry, | |
121 | * and directory contents (for readdir results), or | |
122 | * 2) the file range lock info (for fcntl F_GETLK results). | |
2f2dc053 SW |
123 | */ |
124 | struct ceph_mds_reply_info_parsed { | |
125 | struct ceph_mds_reply_head *head; | |
126 | ||
25933abd | 127 | /* trace */ |
2f2dc053 SW |
128 | struct ceph_mds_reply_info_in diri, targeti; |
129 | struct ceph_mds_reply_dirfrag *dirfrag; | |
130 | char *dname; | |
4ac4c23e | 131 | u8 *altname; |
2f2dc053 | 132 | u32 dname_len; |
4ac4c23e | 133 | u32 altname_len; |
2f2dc053 | 134 | struct ceph_mds_reply_lease *dlease; |
6ddf5f16 | 135 | struct ceph_mds_reply_xattr xattr_info; |
2f2dc053 | 136 | |
25933abd HS |
137 | /* extra */ |
138 | union { | |
139 | /* for fcntl F_GETLK results */ | |
140 | struct ceph_filelock *filelock_reply; | |
141 | ||
142 | /* for readdir results */ | |
143 | struct { | |
144 | struct ceph_mds_reply_dirfrag *dir_dir; | |
54008399 | 145 | size_t dir_buf_size; |
25933abd | 146 | int dir_nr; |
f3c4ebe6 | 147 | bool dir_end; |
79162547 | 148 | bool dir_complete; |
f3c4ebe6 | 149 | bool hash_order; |
79162547 | 150 | bool offset_hash; |
2a5beea3 | 151 | struct ceph_mds_reply_dir_entry *dir_entries; |
25933abd | 152 | }; |
6e8575fa SL |
153 | |
154 | /* for create results */ | |
155 | struct { | |
156 | bool has_create_ino; | |
157 | u64 ino; | |
158 | }; | |
25933abd | 159 | }; |
2f2dc053 SW |
160 | |
161 | /* encoded blob describing snapshot contexts for certain | |
162 | operations (e.g., open) */ | |
163 | void *snapblob; | |
164 | int snapblob_len; | |
165 | }; | |
166 | ||
167 | ||
168 | /* | |
169 | * cap releases are batched and sent to the MDS en masse. | |
92475f05 JL |
170 | * |
171 | * Account for per-message overhead of mds_cap_release header | |
172 | * and __le32 for osd epoch barrier trailing field. | |
2f2dc053 | 173 | */ |
92475f05 | 174 | #define CEPH_CAPS_PER_RELEASE ((PAGE_SIZE - sizeof(u32) - \ |
2f2dc053 | 175 | sizeof(struct ceph_mds_cap_release)) / \ |
92475f05 | 176 | sizeof(struct ceph_mds_cap_item)) |
2f2dc053 SW |
177 | |
178 | ||
179 | /* | |
180 | * state associated with each MDS<->client session | |
181 | */ | |
182 | enum { | |
183 | CEPH_MDS_SESSION_NEW = 1, | |
184 | CEPH_MDS_SESSION_OPENING = 2, | |
185 | CEPH_MDS_SESSION_OPEN = 3, | |
186 | CEPH_MDS_SESSION_HUNG = 4, | |
7e6906c1 YZ |
187 | CEPH_MDS_SESSION_RESTARTING = 5, |
188 | CEPH_MDS_SESSION_RECONNECTING = 6, | |
189 | CEPH_MDS_SESSION_CLOSING = 7, | |
4d681c2f XL |
190 | CEPH_MDS_SESSION_CLOSED = 8, |
191 | CEPH_MDS_SESSION_REJECTED = 9, | |
2f2dc053 SW |
192 | }; |
193 | ||
194 | struct ceph_mds_session { | |
195 | struct ceph_mds_client *s_mdsc; | |
196 | int s_mds; | |
197 | int s_state; | |
198 | unsigned long s_ttl; /* time until mds kills us */ | |
84bf3950 | 199 | unsigned long s_features; |
2f2dc053 SW |
200 | u64 s_seq; /* incoming msg seq # */ |
201 | struct mutex s_mutex; /* serialize session messages */ | |
202 | ||
203 | struct ceph_connection s_con; | |
204 | ||
6c4a1915 | 205 | struct ceph_auth_handshake s_auth; |
4e7a5dcd | 206 | |
52d60f8e JL |
207 | atomic_t s_cap_gen; /* inc each time we get mds stale msg */ |
208 | unsigned long s_cap_ttl; /* when session caps expire. protected by s_mutex */ | |
d8fb02ab AE |
209 | |
210 | /* protected by s_cap_lock */ | |
211 | spinlock_t s_cap_lock; | |
045100cd | 212 | refcount_t s_ref; |
2f2dc053 | 213 | struct list_head s_caps; /* all caps issued by this session */ |
e3ec8d68 | 214 | struct ceph_cap *s_cap_iterator; |
533a2818 | 215 | int s_nr_caps; |
2f2dc053 | 216 | int s_num_cap_releases; |
99a9c273 | 217 | int s_cap_reconnect; |
03f4fcb0 | 218 | int s_readonly; |
2f2dc053 | 219 | struct list_head s_cap_releases; /* waiting cap_release messages */ |
e3ec8d68 | 220 | struct work_struct s_cap_release_work; |
2f2dc053 | 221 | |
829ad4db | 222 | /* See ceph_inode_info->i_dirty_item. */ |
1cf03a68 | 223 | struct list_head s_cap_dirty; /* inodes w/ dirty caps */ |
829ad4db JL |
224 | |
225 | /* See ceph_inode_info->i_flushing_item. */ | |
2f2dc053 | 226 | struct list_head s_cap_flushing; /* inodes w/ flushing caps */ |
1cf03a68 | 227 | |
2f2dc053 SW |
228 | unsigned long s_renew_requested; /* last time we sent a renew req */ |
229 | u64 s_renew_seq; | |
230 | ||
2f2dc053 SW |
231 | struct list_head s_waiting; /* waiting requests */ |
232 | struct list_head s_unsafe; /* unsafe requests */ | |
d4846487 | 233 | struct xarray s_delegated_inos; |
2f2dc053 SW |
234 | }; |
235 | ||
236 | /* | |
237 | * modes of choosing which MDS to send a request to | |
238 | */ | |
239 | enum { | |
240 | USE_ANY_MDS, | |
241 | USE_RANDOM_MDS, | |
242 | USE_AUTH_MDS, /* prefer authoritative mds for this metadata item */ | |
243 | }; | |
244 | ||
245 | struct ceph_mds_request; | |
246 | struct ceph_mds_client; | |
247 | ||
248 | /* | |
249 | * request completion callback | |
250 | */ | |
251 | typedef void (*ceph_mds_request_callback_t) (struct ceph_mds_client *mdsc, | |
252 | struct ceph_mds_request *req); | |
9280be24 YZ |
253 | /* |
254 | * wait for request completion callback | |
255 | */ | |
256 | typedef int (*ceph_mds_request_wait_callback_t) (struct ceph_mds_client *mdsc, | |
257 | struct ceph_mds_request *req); | |
2f2dc053 SW |
258 | |
259 | /* | |
260 | * an in-flight mds request | |
261 | */ | |
262 | struct ceph_mds_request { | |
263 | u64 r_tid; /* transaction id */ | |
44ca18f2 | 264 | struct rb_node r_node; |
37151668 | 265 | struct ceph_mds_client *r_mdsc; |
2f2dc053 | 266 | |
045100cd | 267 | struct kref r_kref; |
2f2dc053 | 268 | int r_op; /* mds op code */ |
2f2dc053 SW |
269 | |
270 | /* operation on what? */ | |
271 | struct inode *r_inode; /* arg1 */ | |
272 | struct dentry *r_dentry; /* arg1 */ | |
273 | struct dentry *r_old_dentry; /* arg2: rename from or link from */ | |
41b02e1f | 274 | struct inode *r_old_dentry_dir; /* arg2: old dentry's parent dir */ |
2f2dc053 SW |
275 | char *r_path1, *r_path2; |
276 | struct ceph_vino r_ino1, r_ino2; | |
277 | ||
3dd69aab | 278 | struct inode *r_parent; /* parent dir inode */ |
2f2dc053 | 279 | struct inode *r_target_inode; /* resulting inode */ |
ec9595c0 | 280 | struct inode *r_new_inode; /* new inode (for creates) */ |
2f2dc053 | 281 | |
bc2de10d JL |
282 | #define CEPH_MDS_R_DIRECT_IS_HASH (1) /* r_direct_hash is valid */ |
283 | #define CEPH_MDS_R_ABORTED (2) /* call was aborted */ | |
284 | #define CEPH_MDS_R_GOT_UNSAFE (3) /* got an unsafe reply */ | |
285 | #define CEPH_MDS_R_GOT_SAFE (4) /* got a safe reply */ | |
286 | #define CEPH_MDS_R_GOT_RESULT (5) /* got a result */ | |
287 | #define CEPH_MDS_R_DID_PREPOPULATE (6) /* prepopulated readdir */ | |
3dd69aab | 288 | #define CEPH_MDS_R_PARENT_LOCKED (7) /* is r_parent->i_rwsem wlocked? */ |
3bb48b41 | 289 | #define CEPH_MDS_R_ASYNC (8) /* async request */ |
16be62fc | 290 | #define CEPH_MDS_R_FSCRYPT_FILE (9) /* must marshal fscrypt_file field */ |
bc2de10d JL |
291 | unsigned long r_req_flags; |
292 | ||
b4556396 SW |
293 | struct mutex r_fill_mutex; |
294 | ||
2f2dc053 | 295 | union ceph_mds_request_args r_args; |
2d332d5b JL |
296 | |
297 | struct ceph_fscrypt_auth *r_fscrypt_auth; | |
16be62fc | 298 | u64 r_fscrypt_file; |
2d332d5b | 299 | |
24865e75 JL |
300 | u8 *r_altname; /* fscrypt binary crypttext for long filenames */ |
301 | u32 r_altname_len; /* length of r_altname */ | |
302 | ||
2f2dc053 | 303 | int r_fmode; /* file mode, if expecting cap */ |
c36d6414 | 304 | int r_request_release_offset; |
9eaa7b79 | 305 | const struct cred *r_cred; |
9c2df227 | 306 | struct mnt_idmap *r_mnt_idmap; |
0ed1e90a | 307 | struct timespec64 r_stamp; |
2f2dc053 SW |
308 | |
309 | /* for choosing which mds to send this request to */ | |
310 | int r_direct_mode; | |
311 | u32 r_direct_hash; /* choose dir frag based on this dentry hash */ | |
2f2dc053 SW |
312 | |
313 | /* data payload is used for xattr ops */ | |
25e6bae3 | 314 | struct ceph_pagelist *r_pagelist; |
2f2dc053 SW |
315 | |
316 | /* what caps shall we drop? */ | |
317 | int r_inode_drop, r_inode_unless; | |
318 | int r_dentry_drop, r_dentry_unless; | |
319 | int r_old_dentry_drop, r_old_dentry_unless; | |
320 | struct inode *r_old_inode; | |
321 | int r_old_inode_drop, r_old_inode_unless; | |
322 | ||
323 | struct ceph_msg *r_request; /* original request */ | |
324 | struct ceph_msg *r_reply; | |
325 | struct ceph_mds_reply_info_parsed r_reply_info; | |
326 | int r_err; | |
9eaa7b79 | 327 | u32 r_readdir_offset; |
a25949b9 JL |
328 | |
329 | struct page *r_locked_page; | |
330 | int r_dir_caps; | |
c36d6414 | 331 | int r_num_caps; |
2f2dc053 | 332 | |
a319bf56 | 333 | unsigned long r_timeout; /* optional. jiffies, 0 is "wait forever" */ |
2f2dc053 | 334 | unsigned long r_started; /* start time to measure timeout against */ |
70c94820 XL |
335 | unsigned long r_start_latency; /* start time to measure latency */ |
336 | unsigned long r_end_latency; /* finish time to measure latency */ | |
2f2dc053 SW |
337 | unsigned long r_request_started; /* start time for mds request only, |
338 | used to measure lease durations */ | |
339 | ||
340 | /* link unsafe requests to parent directory, for fsync */ | |
341 | struct inode *r_unsafe_dir; | |
342 | struct list_head r_unsafe_dir_item; | |
343 | ||
68cd5b4b YZ |
344 | /* unsafe requests that modify the target inode */ |
345 | struct list_head r_unsafe_target_item; | |
346 | ||
2f2dc053 SW |
347 | struct ceph_mds_session *r_session; |
348 | ||
349 | int r_attempts; /* resend attempts */ | |
350 | int r_num_fwd; /* number of forward attempts */ | |
2f2dc053 | 351 | int r_resend_mds; /* mds to resend to next, if any*/ |
e55b71f8 | 352 | u32 r_sent_on_mseq; /* cap mseq request was sent at*/ |
6deb8008 | 353 | u64 r_deleg_ino; |
2f2dc053 | 354 | |
2f2dc053 SW |
355 | struct list_head r_wait; |
356 | struct completion r_completion; | |
357 | struct completion r_safe_completion; | |
358 | ceph_mds_request_callback_t r_callback; | |
359 | struct list_head r_unsafe_item; /* per-session unsafe list item */ | |
2f2dc053 | 360 | |
fdd4e158 YZ |
361 | long long r_dir_release_cnt; |
362 | long long r_dir_ordered_cnt; | |
363 | int r_readdir_cache_idx; | |
2f2dc053 | 364 | |
6eb06c46 XL |
365 | int r_feature_needed; |
366 | ||
2f2dc053 | 367 | struct ceph_cap_reservation r_caps_reservation; |
2f2dc053 SW |
368 | }; |
369 | ||
10183a69 YZ |
370 | struct ceph_pool_perm { |
371 | struct rb_node node; | |
10183a69 | 372 | int perm; |
7627151e | 373 | s64 pool; |
779fe0fb YZ |
374 | size_t pool_ns_len; |
375 | char pool_ns[]; | |
10183a69 YZ |
376 | }; |
377 | ||
75c9627e YZ |
378 | struct ceph_snapid_map { |
379 | struct rb_node node; | |
380 | struct list_head lru; | |
381 | atomic_t ref; | |
75c9627e | 382 | dev_t dev; |
db2993a4 | 383 | u64 snap; |
75c9627e YZ |
384 | unsigned long last_used; |
385 | }; | |
386 | ||
0c44a8e0 LH |
387 | /* |
388 | * node for list of quotarealm inodes that are not visible from the filesystem | |
389 | * mountpoint, but required to handle, e.g. quotas. | |
390 | */ | |
391 | struct ceph_quotarealm_inode { | |
392 | struct rb_node node; | |
393 | u64 ino; | |
394 | unsigned long timeout; /* last time a lookup failed for this inode */ | |
395 | struct mutex mutex; | |
396 | struct inode *inode; | |
397 | }; | |
398 | ||
3a3430af JL |
399 | struct cap_wait { |
400 | struct list_head list; | |
ebce3eb2 | 401 | u64 ino; |
3a3430af JL |
402 | pid_t tgid; |
403 | int need; | |
404 | int want; | |
405 | }; | |
406 | ||
e7e607bd | 407 | enum { |
e3dfcab2 XL |
408 | CEPH_MDSC_STOPPING_BEGIN = 1, |
409 | CEPH_MDSC_STOPPING_FLUSHING = 2, | |
410 | CEPH_MDSC_STOPPING_FLUSHED = 3, | |
e7e607bd XL |
411 | }; |
412 | ||
2f2dc053 SW |
413 | /* |
414 | * mds client state | |
415 | */ | |
416 | struct ceph_mds_client { | |
3d14c5d2 | 417 | struct ceph_fs_client *fsc; |
2f2dc053 SW |
418 | struct mutex mutex; /* all nested structures */ |
419 | ||
420 | struct ceph_mdsmap *mdsmap; | |
f3c60c59 SW |
421 | struct completion safe_umount_waiters; |
422 | wait_queue_head_t session_close_wq; | |
2f2dc053 | 423 | struct list_head waiting_for_map; |
430afbad | 424 | int mdsmap_err; |
2f2dc053 SW |
425 | |
426 | struct ceph_mds_session **sessions; /* NULL for mds if no session */ | |
86d8f67b | 427 | atomic_t num_sessions; |
c74d79af | 428 | int max_sessions; /* len of sessions array */ |
e3dfcab2 XL |
429 | |
430 | spinlock_t stopping_lock; /* protect snap_empty */ | |
431 | int stopping; /* the stage of shutting down */ | |
432 | atomic_t stopping_blockers; | |
433 | struct completion stopping_waiter; | |
2f2dc053 | 434 | |
d557c48d | 435 | atomic64_t quotarealms_count; /* # realms with quota */ |
0c44a8e0 LH |
436 | /* |
437 | * We keep a list of inodes we don't see in the mountpoint but that we | |
438 | * need to track quota realms. | |
439 | */ | |
440 | struct rb_root quotarealms_inodes; | |
441 | struct mutex quotarealms_inodes_mutex; | |
d557c48d | 442 | |
2f2dc053 SW |
443 | /* |
444 | * snap_rwsem will cover cap linkage into snaprealms, and | |
445 | * realm snap contexts. (later, we can do per-realm snap | |
446 | * contexts locks..) the empty list contains realms with no | |
447 | * references (implying they contain no inodes with caps) that | |
448 | * should be destroyed. | |
449 | */ | |
affbc19a | 450 | u64 last_snap_seq; |
2f2dc053 | 451 | struct rw_semaphore snap_rwsem; |
a105f00c | 452 | struct rb_root snap_realms; |
2f2dc053 | 453 | struct list_head snap_empty; |
81c5a148 | 454 | int num_snap_realms; |
2f2dc053 SW |
455 | spinlock_t snap_empty_lock; /* protect snap_empty */ |
456 | ||
457 | u64 last_tid; /* most recent mds request */ | |
e8a7b8b1 YZ |
458 | u64 oldest_tid; /* oldest incomplete mds request, |
459 | excluding setfilelock requests */ | |
44ca18f2 | 460 | struct rb_root request_tree; /* pending mds requests */ |
2f2dc053 SW |
461 | struct delayed_work delayed_work; /* delayed work */ |
462 | unsigned long last_renew_caps; /* last time we renewed our caps */ | |
463 | struct list_head cap_delay_list; /* caps with delayed release */ | |
464 | spinlock_t cap_delay_lock; /* protects cap_delay_list */ | |
465 | struct list_head snap_flush_list; /* cap_snaps ready to flush */ | |
466 | spinlock_t snap_flush_lock; | |
467 | ||
553adfd9 | 468 | u64 last_cap_flush_tid; |
e4500b5e | 469 | struct list_head cap_flush_list; |
db354052 | 470 | struct list_head cap_dirty_migrating; /* ...that are migration... */ |
2f2dc053 SW |
471 | int num_cap_flushing; /* # caps we are flushing */ |
472 | spinlock_t cap_dirty_lock; /* protects above items */ | |
473 | wait_queue_head_t cap_flushing_wq; | |
474 | ||
37c4efc1 | 475 | struct work_struct cap_reclaim_work; |
fe33032d | 476 | atomic_t cap_reclaim_pending; |
37c4efc1 | 477 | |
37151668 YS |
478 | /* |
479 | * Cap reservations | |
480 | * | |
481 | * Maintain a global pool of preallocated struct ceph_caps, referenced | |
482 | * by struct ceph_caps_reservations. This ensures that we preallocate | |
483 | * memory needed to successfully process an MDS response. (If an MDS | |
484 | * sends us cap information and we fail to process it, we will have | |
485 | * problems due to the client and MDS being out of sync.) | |
486 | * | |
487 | * Reservations are 'owned' by a ceph_cap_reservation context. | |
488 | */ | |
489 | spinlock_t caps_list_lock; | |
490 | struct list_head caps_list; /* unused (reserved or | |
491 | unreserved) */ | |
3a3430af | 492 | struct list_head cap_wait_list; |
37151668 YS |
493 | int caps_total_count; /* total caps allocated */ |
494 | int caps_use_count; /* in use */ | |
fe33032d | 495 | int caps_use_max; /* max used caps */ |
37151668 YS |
496 | int caps_reserve_count; /* unused, reserved */ |
497 | int caps_avail_count; /* unused, unreserved */ | |
498 | int caps_min_count; /* keep at least this many | |
499 | (unreserved) */ | |
37c4efc1 YZ |
500 | spinlock_t dentry_list_lock; |
501 | struct list_head dentry_leases; /* fifo list */ | |
502 | struct list_head dentry_dir_leases; /* lru list */ | |
10183a69 | 503 | |
f9009efa XL |
504 | struct ceph_client_metric metric; |
505 | ||
75c9627e YZ |
506 | spinlock_t snapid_map_lock; |
507 | struct rb_root snapid_map_tree; | |
508 | struct list_head snapid_map_lru; | |
509 | ||
10183a69 YZ |
510 | struct rw_semaphore pool_perm_rwsem; |
511 | struct rb_root pool_perm_tree; | |
717e6f28 YZ |
512 | |
513 | char nodename[__NEW_UTS_LEN + 1]; | |
2f2dc053 SW |
514 | }; |
515 | ||
516 | extern const char *ceph_mds_op_name(int op); | |
517 | ||
3e699bd8 | 518 | extern bool check_session_state(struct ceph_mds_session *s); |
62575e27 | 519 | void inc_session_sequence(struct ceph_mds_session *s); |
3e699bd8 | 520 | |
2f2dc053 SW |
521 | extern struct ceph_mds_session * |
522 | __ceph_lookup_mds_session(struct ceph_mds_client *, int mds); | |
523 | ||
a687ecaf JS |
524 | extern const char *ceph_session_state_name(int s); |
525 | ||
5b3248c6 XL |
526 | extern struct ceph_mds_session * |
527 | ceph_get_mds_session(struct ceph_mds_session *s); | |
2f2dc053 SW |
528 | extern void ceph_put_mds_session(struct ceph_mds_session *s); |
529 | ||
530 | extern int ceph_send_msg_mds(struct ceph_mds_client *mdsc, | |
531 | struct ceph_msg *msg, int mds); | |
532 | ||
3d14c5d2 | 533 | extern int ceph_mdsc_init(struct ceph_fs_client *fsc); |
2f2dc053 | 534 | extern void ceph_mdsc_close_sessions(struct ceph_mds_client *mdsc); |
48fec5d0 | 535 | extern void ceph_mdsc_force_umount(struct ceph_mds_client *mdsc); |
3d14c5d2 | 536 | extern void ceph_mdsc_destroy(struct ceph_fs_client *fsc); |
2f2dc053 SW |
537 | |
538 | extern void ceph_mdsc_sync(struct ceph_mds_client *mdsc); | |
539 | ||
167c9e35 | 540 | extern void ceph_invalidate_dir_request(struct ceph_mds_request *req); |
54008399 YZ |
541 | extern int ceph_alloc_readdir_reply_buffer(struct ceph_mds_request *req, |
542 | struct inode *dir); | |
2f2dc053 SW |
543 | extern struct ceph_mds_request * |
544 | ceph_mdsc_create_request(struct ceph_mds_client *mdsc, int op, int mode); | |
86bda539 JL |
545 | extern int ceph_mdsc_submit_request(struct ceph_mds_client *mdsc, |
546 | struct inode *dir, | |
547 | struct ceph_mds_request *req); | |
9eaa7b79 JL |
548 | int ceph_mdsc_wait_request(struct ceph_mds_client *mdsc, |
549 | struct ceph_mds_request *req, | |
550 | ceph_mds_request_wait_callback_t wait_func); | |
2f2dc053 SW |
551 | extern int ceph_mdsc_do_request(struct ceph_mds_client *mdsc, |
552 | struct inode *dir, | |
553 | struct ceph_mds_request *req); | |
a25949b9 | 554 | extern void ceph_mdsc_release_dir_caps(struct ceph_mds_request *req); |
07045648 | 555 | extern void ceph_mdsc_release_dir_caps_async(struct ceph_mds_request *req); |
2f2dc053 SW |
556 | static inline void ceph_mdsc_get_request(struct ceph_mds_request *req) |
557 | { | |
153c8e6b SW |
558 | kref_get(&req->r_kref); |
559 | } | |
560 | extern void ceph_mdsc_release_request(struct kref *kref); | |
561 | static inline void ceph_mdsc_put_request(struct ceph_mds_request *req) | |
562 | { | |
563 | kref_put(&req->r_kref, ceph_mdsc_release_request); | |
2f2dc053 | 564 | } |
2f2dc053 | 565 | |
d095559c | 566 | extern void send_flush_mdlog(struct ceph_mds_session *s); |
59b312f3 XL |
567 | extern void ceph_mdsc_iterate_sessions(struct ceph_mds_client *mdsc, |
568 | void (*cb)(struct ceph_mds_session *), | |
569 | bool check_state); | |
fba97e80 | 570 | extern struct ceph_msg *ceph_create_session_msg(u32 op, u64 seq); |
e3ec8d68 YZ |
571 | extern void __ceph_queue_cap_release(struct ceph_mds_session *session, |
572 | struct ceph_cap *cap); | |
573 | extern void ceph_flush_cap_releases(struct ceph_mds_client *mdsc, | |
574 | struct ceph_mds_session *session); | |
37c4efc1 | 575 | extern void ceph_queue_cap_reclaim_work(struct ceph_mds_client *mdsc); |
fe33032d | 576 | extern void ceph_reclaim_caps_nr(struct ceph_mds_client *mdsc, int nr); |
f5d77269 | 577 | extern int ceph_iterate_session_caps(struct ceph_mds_session *session, |
aaf67de7 | 578 | int (*cb)(struct inode *, int mds, void *), |
f5d77269 | 579 | void *arg); |
2f2dc053 SW |
580 | extern void ceph_mdsc_pre_umount(struct ceph_mds_client *mdsc); |
581 | ||
f77f21bb JL |
582 | static inline void ceph_mdsc_free_path(char *path, int len) |
583 | { | |
2a575f13 | 584 | if (!IS_ERR_OR_NULL(path)) |
f77f21bb JL |
585 | __putname(path - (PATH_MAX - 1 - len)); |
586 | } | |
587 | ||
197b7d79 XL |
588 | extern char *ceph_mdsc_build_path(struct ceph_mds_client *mdsc, |
589 | struct dentry *dentry, int *plen, u64 *base, | |
3fd945a7 | 590 | int for_wire); |
2f2dc053 SW |
591 | |
592 | extern void __ceph_mdsc_drop_dentry_lease(struct dentry *dentry); | |
593 | extern void ceph_mdsc_lease_send_msg(struct ceph_mds_session *session, | |
2f2dc053 SW |
594 | struct dentry *dentry, char action, |
595 | u32 seq); | |
596 | ||
430afbad YZ |
597 | extern void ceph_mdsc_handle_mdsmap(struct ceph_mds_client *mdsc, |
598 | struct ceph_msg *msg); | |
599 | extern void ceph_mdsc_handle_fsmap(struct ceph_mds_client *mdsc, | |
600 | struct ceph_msg *msg); | |
2f2dc053 | 601 | |
5d72d13c YZ |
602 | extern struct ceph_mds_session * |
603 | ceph_mdsc_open_export_target_session(struct ceph_mds_client *mdsc, int target); | |
154f42c2 SW |
604 | extern void ceph_mdsc_open_export_target_sessions(struct ceph_mds_client *mdsc, |
605 | struct ceph_mds_session *session); | |
606 | ||
e30ee581 ZZ |
607 | extern int ceph_trim_caps(struct ceph_mds_client *mdsc, |
608 | struct ceph_mds_session *session, | |
609 | int max_caps); | |
d4846487 | 610 | |
891f3f5a JL |
611 | static inline int ceph_wait_on_async_create(struct inode *inode) |
612 | { | |
613 | struct ceph_inode_info *ci = ceph_inode(inode); | |
614 | ||
615 | return wait_on_bit(&ci->i_ceph_flags, CEPH_ASYNC_CREATE_BIT, | |
5e56776d | 616 | TASK_KILLABLE); |
891f3f5a | 617 | } |
d4846487 | 618 | |
4868e537 | 619 | extern int ceph_wait_on_conflict_unlink(struct dentry *dentry); |
d4846487 JL |
620 | extern u64 ceph_get_deleg_ino(struct ceph_mds_session *session); |
621 | extern int ceph_restore_deleg_ino(struct ceph_mds_session *session, u64 ino); | |
673478b6 AM |
622 | |
623 | extern bool enable_unsafe_idmap; | |
2f2dc053 | 624 | #endif |