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> | |
17 | #include <linux/ceph/mdsmap.h> | |
6c4a1915 | 18 | #include <linux/ceph/auth.h> |
2f2dc053 | 19 | |
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, |
d2f8bb27 | 35 | |
6eb06c46 | 36 | CEPHFS_FEATURE_MAX = CEPHFS_FEATURE_OP_GETVXATTR, |
9ba1e224 XL |
37 | }; |
38 | ||
9ba1e224 | 39 | #define CEPHFS_FEATURES_CLIENT_SUPPORTED { \ |
d2f8bb27 YZ |
40 | 0, 1, 2, 3, 4, 5, 6, 7, \ |
41 | CEPHFS_FEATURE_MIMIC, \ | |
b37fe1f9 | 42 | CEPHFS_FEATURE_REPLY_ENCODING, \ |
d2f8bb27 | 43 | CEPHFS_FEATURE_LAZY_CAP_WANTED, \ |
81c5a148 | 44 | CEPHFS_FEATURE_MULTI_RECONNECT, \ |
d4846487 | 45 | CEPHFS_FEATURE_DELEG_INO, \ |
18f473b3 | 46 | CEPHFS_FEATURE_METRIC_COLLECT, \ |
4ac4c23e | 47 | CEPHFS_FEATURE_ALTERNATE_NAME, \ |
300e42a2 | 48 | CEPHFS_FEATURE_NOTIFY_SESSION_STATE, \ |
6eb06c46 | 49 | CEPHFS_FEATURE_OP_GETVXATTR, \ |
342ce182 | 50 | } |
342ce182 | 51 | |
2f2dc053 SW |
52 | /* |
53 | * Some lock dependencies: | |
54 | * | |
55 | * session->s_mutex | |
56 | * mdsc->mutex | |
57 | * | |
58 | * mdsc->snap_rwsem | |
59 | * | |
be655596 | 60 | * ci->i_ceph_lock |
2f2dc053 SW |
61 | * mdsc->snap_flush_lock |
62 | * mdsc->cap_delay_lock | |
63 | * | |
64 | */ | |
65 | ||
3d14c5d2 | 66 | struct ceph_fs_client; |
2f2dc053 SW |
67 | struct ceph_cap; |
68 | ||
69 | /* | |
70 | * parsed info about a single inode. pointers are into the encoded | |
71 | * on-wire structures within the mds reply message payload. | |
72 | */ | |
73 | struct ceph_mds_reply_info_in { | |
74 | struct ceph_mds_reply_inode *in; | |
14303d20 | 75 | struct ceph_dir_layout dir_layout; |
2f2dc053 SW |
76 | u32 symlink_len; |
77 | char *symlink; | |
78 | u32 xattr_len; | |
79 | char *xattr_data; | |
fb01d1f8 YZ |
80 | u64 inline_version; |
81 | u32 inline_len; | |
82 | char *inline_data; | |
5ea5c5e0 | 83 | u32 pool_ns_len; |
779fe0fb | 84 | char *pool_ns_data; |
fb18a575 LH |
85 | u64 max_bytes; |
86 | u64 max_files; | |
08796873 | 87 | s32 dir_pin; |
245ce991 | 88 | struct ceph_timespec btime; |
193e7b37 | 89 | struct ceph_timespec snap_btime; |
2d332d5b JL |
90 | u8 *fscrypt_auth; |
91 | u8 *fscrypt_file; | |
92 | u32 fscrypt_auth_len; | |
93 | u32 fscrypt_file_len; | |
e7f72952 | 94 | u64 rsnaps; |
a35ead31 | 95 | u64 change_attr; |
2f2dc053 SW |
96 | }; |
97 | ||
2a5beea3 YZ |
98 | struct ceph_mds_reply_dir_entry { |
99 | char *name; | |
4ac4c23e | 100 | u8 *altname; |
2a5beea3 | 101 | u32 name_len; |
4ac4c23e | 102 | u32 altname_len; |
2a5beea3 YZ |
103 | struct ceph_mds_reply_lease *lease; |
104 | struct ceph_mds_reply_info_in inode; | |
8974eebd | 105 | loff_t offset; |
2a5beea3 YZ |
106 | }; |
107 | ||
6ddf5f16 MC |
108 | struct ceph_mds_reply_xattr { |
109 | char *xattr_value; | |
110 | size_t xattr_value_len; | |
111 | }; | |
112 | ||
2f2dc053 | 113 | /* |
25933abd HS |
114 | * parsed info about an mds reply, including information about |
115 | * either: 1) the target inode and/or its parent directory and dentry, | |
116 | * and directory contents (for readdir results), or | |
117 | * 2) the file range lock info (for fcntl F_GETLK results). | |
2f2dc053 SW |
118 | */ |
119 | struct ceph_mds_reply_info_parsed { | |
120 | struct ceph_mds_reply_head *head; | |
121 | ||
25933abd | 122 | /* trace */ |
2f2dc053 SW |
123 | struct ceph_mds_reply_info_in diri, targeti; |
124 | struct ceph_mds_reply_dirfrag *dirfrag; | |
125 | char *dname; | |
4ac4c23e | 126 | u8 *altname; |
2f2dc053 | 127 | u32 dname_len; |
4ac4c23e | 128 | u32 altname_len; |
2f2dc053 | 129 | struct ceph_mds_reply_lease *dlease; |
6ddf5f16 | 130 | struct ceph_mds_reply_xattr xattr_info; |
2f2dc053 | 131 | |
25933abd HS |
132 | /* extra */ |
133 | union { | |
134 | /* for fcntl F_GETLK results */ | |
135 | struct ceph_filelock *filelock_reply; | |
136 | ||
137 | /* for readdir results */ | |
138 | struct { | |
139 | struct ceph_mds_reply_dirfrag *dir_dir; | |
54008399 | 140 | size_t dir_buf_size; |
25933abd | 141 | int dir_nr; |
f3c4ebe6 | 142 | bool dir_end; |
79162547 | 143 | bool dir_complete; |
f3c4ebe6 | 144 | bool hash_order; |
79162547 | 145 | bool offset_hash; |
2a5beea3 | 146 | struct ceph_mds_reply_dir_entry *dir_entries; |
25933abd | 147 | }; |
6e8575fa SL |
148 | |
149 | /* for create results */ | |
150 | struct { | |
151 | bool has_create_ino; | |
152 | u64 ino; | |
153 | }; | |
25933abd | 154 | }; |
2f2dc053 SW |
155 | |
156 | /* encoded blob describing snapshot contexts for certain | |
157 | operations (e.g., open) */ | |
158 | void *snapblob; | |
159 | int snapblob_len; | |
160 | }; | |
161 | ||
162 | ||
163 | /* | |
164 | * cap releases are batched and sent to the MDS en masse. | |
92475f05 JL |
165 | * |
166 | * Account for per-message overhead of mds_cap_release header | |
167 | * and __le32 for osd epoch barrier trailing field. | |
2f2dc053 | 168 | */ |
92475f05 | 169 | #define CEPH_CAPS_PER_RELEASE ((PAGE_SIZE - sizeof(u32) - \ |
2f2dc053 | 170 | sizeof(struct ceph_mds_cap_release)) / \ |
92475f05 | 171 | sizeof(struct ceph_mds_cap_item)) |
2f2dc053 SW |
172 | |
173 | ||
174 | /* | |
175 | * state associated with each MDS<->client session | |
176 | */ | |
177 | enum { | |
178 | CEPH_MDS_SESSION_NEW = 1, | |
179 | CEPH_MDS_SESSION_OPENING = 2, | |
180 | CEPH_MDS_SESSION_OPEN = 3, | |
181 | CEPH_MDS_SESSION_HUNG = 4, | |
7e6906c1 YZ |
182 | CEPH_MDS_SESSION_RESTARTING = 5, |
183 | CEPH_MDS_SESSION_RECONNECTING = 6, | |
184 | CEPH_MDS_SESSION_CLOSING = 7, | |
4d681c2f XL |
185 | CEPH_MDS_SESSION_CLOSED = 8, |
186 | CEPH_MDS_SESSION_REJECTED = 9, | |
2f2dc053 SW |
187 | }; |
188 | ||
189 | struct ceph_mds_session { | |
190 | struct ceph_mds_client *s_mdsc; | |
191 | int s_mds; | |
192 | int s_state; | |
193 | unsigned long s_ttl; /* time until mds kills us */ | |
84bf3950 | 194 | unsigned long s_features; |
2f2dc053 SW |
195 | u64 s_seq; /* incoming msg seq # */ |
196 | struct mutex s_mutex; /* serialize session messages */ | |
197 | ||
198 | struct ceph_connection s_con; | |
199 | ||
6c4a1915 | 200 | struct ceph_auth_handshake s_auth; |
4e7a5dcd | 201 | |
52d60f8e JL |
202 | atomic_t s_cap_gen; /* inc each time we get mds stale msg */ |
203 | unsigned long s_cap_ttl; /* when session caps expire. protected by s_mutex */ | |
d8fb02ab AE |
204 | |
205 | /* protected by s_cap_lock */ | |
206 | spinlock_t s_cap_lock; | |
045100cd | 207 | refcount_t s_ref; |
2f2dc053 | 208 | struct list_head s_caps; /* all caps issued by this session */ |
e3ec8d68 | 209 | struct ceph_cap *s_cap_iterator; |
533a2818 | 210 | int s_nr_caps; |
2f2dc053 | 211 | int s_num_cap_releases; |
99a9c273 | 212 | int s_cap_reconnect; |
03f4fcb0 | 213 | int s_readonly; |
2f2dc053 | 214 | struct list_head s_cap_releases; /* waiting cap_release messages */ |
e3ec8d68 | 215 | struct work_struct s_cap_release_work; |
2f2dc053 | 216 | |
829ad4db | 217 | /* See ceph_inode_info->i_dirty_item. */ |
1cf03a68 | 218 | struct list_head s_cap_dirty; /* inodes w/ dirty caps */ |
829ad4db JL |
219 | |
220 | /* See ceph_inode_info->i_flushing_item. */ | |
2f2dc053 | 221 | struct list_head s_cap_flushing; /* inodes w/ flushing caps */ |
1cf03a68 | 222 | |
2f2dc053 SW |
223 | unsigned long s_renew_requested; /* last time we sent a renew req */ |
224 | u64 s_renew_seq; | |
225 | ||
2f2dc053 SW |
226 | struct list_head s_waiting; /* waiting requests */ |
227 | struct list_head s_unsafe; /* unsafe requests */ | |
d4846487 | 228 | struct xarray s_delegated_inos; |
2f2dc053 SW |
229 | }; |
230 | ||
231 | /* | |
232 | * modes of choosing which MDS to send a request to | |
233 | */ | |
234 | enum { | |
235 | USE_ANY_MDS, | |
236 | USE_RANDOM_MDS, | |
237 | USE_AUTH_MDS, /* prefer authoritative mds for this metadata item */ | |
238 | }; | |
239 | ||
240 | struct ceph_mds_request; | |
241 | struct ceph_mds_client; | |
242 | ||
243 | /* | |
244 | * request completion callback | |
245 | */ | |
246 | typedef void (*ceph_mds_request_callback_t) (struct ceph_mds_client *mdsc, | |
247 | struct ceph_mds_request *req); | |
9280be24 YZ |
248 | /* |
249 | * wait for request completion callback | |
250 | */ | |
251 | typedef int (*ceph_mds_request_wait_callback_t) (struct ceph_mds_client *mdsc, | |
252 | struct ceph_mds_request *req); | |
2f2dc053 SW |
253 | |
254 | /* | |
255 | * an in-flight mds request | |
256 | */ | |
257 | struct ceph_mds_request { | |
258 | u64 r_tid; /* transaction id */ | |
44ca18f2 | 259 | struct rb_node r_node; |
37151668 | 260 | struct ceph_mds_client *r_mdsc; |
2f2dc053 | 261 | |
045100cd | 262 | struct kref r_kref; |
2f2dc053 | 263 | int r_op; /* mds op code */ |
2f2dc053 SW |
264 | |
265 | /* operation on what? */ | |
266 | struct inode *r_inode; /* arg1 */ | |
267 | struct dentry *r_dentry; /* arg1 */ | |
268 | struct dentry *r_old_dentry; /* arg2: rename from or link from */ | |
41b02e1f | 269 | struct inode *r_old_dentry_dir; /* arg2: old dentry's parent dir */ |
2f2dc053 SW |
270 | char *r_path1, *r_path2; |
271 | struct ceph_vino r_ino1, r_ino2; | |
272 | ||
3dd69aab | 273 | struct inode *r_parent; /* parent dir inode */ |
2f2dc053 | 274 | struct inode *r_target_inode; /* resulting inode */ |
ec9595c0 | 275 | struct inode *r_new_inode; /* new inode (for creates) */ |
2f2dc053 | 276 | |
bc2de10d JL |
277 | #define CEPH_MDS_R_DIRECT_IS_HASH (1) /* r_direct_hash is valid */ |
278 | #define CEPH_MDS_R_ABORTED (2) /* call was aborted */ | |
279 | #define CEPH_MDS_R_GOT_UNSAFE (3) /* got an unsafe reply */ | |
280 | #define CEPH_MDS_R_GOT_SAFE (4) /* got a safe reply */ | |
281 | #define CEPH_MDS_R_GOT_RESULT (5) /* got a result */ | |
282 | #define CEPH_MDS_R_DID_PREPOPULATE (6) /* prepopulated readdir */ | |
3dd69aab | 283 | #define CEPH_MDS_R_PARENT_LOCKED (7) /* is r_parent->i_rwsem wlocked? */ |
3bb48b41 | 284 | #define CEPH_MDS_R_ASYNC (8) /* async request */ |
bc2de10d JL |
285 | unsigned long r_req_flags; |
286 | ||
b4556396 SW |
287 | struct mutex r_fill_mutex; |
288 | ||
2f2dc053 | 289 | union ceph_mds_request_args r_args; |
2d332d5b JL |
290 | |
291 | struct ceph_fscrypt_auth *r_fscrypt_auth; | |
292 | ||
24865e75 JL |
293 | u8 *r_altname; /* fscrypt binary crypttext for long filenames */ |
294 | u32 r_altname_len; /* length of r_altname */ | |
295 | ||
2f2dc053 | 296 | int r_fmode; /* file mode, if expecting cap */ |
c36d6414 | 297 | int r_request_release_offset; |
9eaa7b79 | 298 | const struct cred *r_cred; |
0ed1e90a | 299 | struct timespec64 r_stamp; |
2f2dc053 SW |
300 | |
301 | /* for choosing which mds to send this request to */ | |
302 | int r_direct_mode; | |
303 | u32 r_direct_hash; /* choose dir frag based on this dentry hash */ | |
2f2dc053 SW |
304 | |
305 | /* data payload is used for xattr ops */ | |
25e6bae3 | 306 | struct ceph_pagelist *r_pagelist; |
2f2dc053 SW |
307 | |
308 | /* what caps shall we drop? */ | |
309 | int r_inode_drop, r_inode_unless; | |
310 | int r_dentry_drop, r_dentry_unless; | |
311 | int r_old_dentry_drop, r_old_dentry_unless; | |
312 | struct inode *r_old_inode; | |
313 | int r_old_inode_drop, r_old_inode_unless; | |
314 | ||
315 | struct ceph_msg *r_request; /* original request */ | |
316 | struct ceph_msg *r_reply; | |
317 | struct ceph_mds_reply_info_parsed r_reply_info; | |
318 | int r_err; | |
9eaa7b79 | 319 | u32 r_readdir_offset; |
a25949b9 JL |
320 | |
321 | struct page *r_locked_page; | |
322 | int r_dir_caps; | |
c36d6414 | 323 | int r_num_caps; |
2f2dc053 | 324 | |
a319bf56 | 325 | unsigned long r_timeout; /* optional. jiffies, 0 is "wait forever" */ |
2f2dc053 | 326 | unsigned long r_started; /* start time to measure timeout against */ |
70c94820 XL |
327 | unsigned long r_start_latency; /* start time to measure latency */ |
328 | unsigned long r_end_latency; /* finish time to measure latency */ | |
2f2dc053 SW |
329 | unsigned long r_request_started; /* start time for mds request only, |
330 | used to measure lease durations */ | |
331 | ||
332 | /* link unsafe requests to parent directory, for fsync */ | |
333 | struct inode *r_unsafe_dir; | |
334 | struct list_head r_unsafe_dir_item; | |
335 | ||
68cd5b4b YZ |
336 | /* unsafe requests that modify the target inode */ |
337 | struct list_head r_unsafe_target_item; | |
338 | ||
2f2dc053 SW |
339 | struct ceph_mds_session *r_session; |
340 | ||
341 | int r_attempts; /* resend attempts */ | |
342 | int r_num_fwd; /* number of forward attempts */ | |
2f2dc053 | 343 | int r_resend_mds; /* mds to resend to next, if any*/ |
e55b71f8 | 344 | u32 r_sent_on_mseq; /* cap mseq request was sent at*/ |
6deb8008 | 345 | u64 r_deleg_ino; |
2f2dc053 | 346 | |
2f2dc053 SW |
347 | struct list_head r_wait; |
348 | struct completion r_completion; | |
349 | struct completion r_safe_completion; | |
350 | ceph_mds_request_callback_t r_callback; | |
351 | struct list_head r_unsafe_item; /* per-session unsafe list item */ | |
2f2dc053 | 352 | |
fdd4e158 YZ |
353 | long long r_dir_release_cnt; |
354 | long long r_dir_ordered_cnt; | |
355 | int r_readdir_cache_idx; | |
2f2dc053 | 356 | |
6eb06c46 XL |
357 | int r_feature_needed; |
358 | ||
2f2dc053 | 359 | struct ceph_cap_reservation r_caps_reservation; |
2f2dc053 SW |
360 | }; |
361 | ||
10183a69 YZ |
362 | struct ceph_pool_perm { |
363 | struct rb_node node; | |
10183a69 | 364 | int perm; |
7627151e | 365 | s64 pool; |
779fe0fb YZ |
366 | size_t pool_ns_len; |
367 | char pool_ns[]; | |
10183a69 YZ |
368 | }; |
369 | ||
75c9627e YZ |
370 | struct ceph_snapid_map { |
371 | struct rb_node node; | |
372 | struct list_head lru; | |
373 | atomic_t ref; | |
75c9627e | 374 | dev_t dev; |
db2993a4 | 375 | u64 snap; |
75c9627e YZ |
376 | unsigned long last_used; |
377 | }; | |
378 | ||
0c44a8e0 LH |
379 | /* |
380 | * node for list of quotarealm inodes that are not visible from the filesystem | |
381 | * mountpoint, but required to handle, e.g. quotas. | |
382 | */ | |
383 | struct ceph_quotarealm_inode { | |
384 | struct rb_node node; | |
385 | u64 ino; | |
386 | unsigned long timeout; /* last time a lookup failed for this inode */ | |
387 | struct mutex mutex; | |
388 | struct inode *inode; | |
389 | }; | |
390 | ||
3a3430af JL |
391 | struct cap_wait { |
392 | struct list_head list; | |
ebce3eb2 | 393 | u64 ino; |
3a3430af JL |
394 | pid_t tgid; |
395 | int need; | |
396 | int want; | |
397 | }; | |
398 | ||
e7e607bd XL |
399 | enum { |
400 | CEPH_MDSC_STOPPING_BEGIN = 1, | |
401 | CEPH_MDSC_STOPPING_FLUSHED = 2, | |
402 | }; | |
403 | ||
2f2dc053 SW |
404 | /* |
405 | * mds client state | |
406 | */ | |
407 | struct ceph_mds_client { | |
3d14c5d2 | 408 | struct ceph_fs_client *fsc; |
2f2dc053 SW |
409 | struct mutex mutex; /* all nested structures */ |
410 | ||
411 | struct ceph_mdsmap *mdsmap; | |
f3c60c59 SW |
412 | struct completion safe_umount_waiters; |
413 | wait_queue_head_t session_close_wq; | |
2f2dc053 | 414 | struct list_head waiting_for_map; |
430afbad | 415 | int mdsmap_err; |
2f2dc053 SW |
416 | |
417 | struct ceph_mds_session **sessions; /* NULL for mds if no session */ | |
86d8f67b | 418 | atomic_t num_sessions; |
c74d79af | 419 | int max_sessions; /* len of sessions array */ |
2f2dc053 SW |
420 | int stopping; /* true if shutting down */ |
421 | ||
d557c48d | 422 | atomic64_t quotarealms_count; /* # realms with quota */ |
0c44a8e0 LH |
423 | /* |
424 | * We keep a list of inodes we don't see in the mountpoint but that we | |
425 | * need to track quota realms. | |
426 | */ | |
427 | struct rb_root quotarealms_inodes; | |
428 | struct mutex quotarealms_inodes_mutex; | |
d557c48d | 429 | |
2f2dc053 SW |
430 | /* |
431 | * snap_rwsem will cover cap linkage into snaprealms, and | |
432 | * realm snap contexts. (later, we can do per-realm snap | |
433 | * contexts locks..) the empty list contains realms with no | |
434 | * references (implying they contain no inodes with caps) that | |
435 | * should be destroyed. | |
436 | */ | |
affbc19a | 437 | u64 last_snap_seq; |
2f2dc053 | 438 | struct rw_semaphore snap_rwsem; |
a105f00c | 439 | struct rb_root snap_realms; |
2f2dc053 | 440 | struct list_head snap_empty; |
81c5a148 | 441 | int num_snap_realms; |
2f2dc053 SW |
442 | spinlock_t snap_empty_lock; /* protect snap_empty */ |
443 | ||
444 | u64 last_tid; /* most recent mds request */ | |
e8a7b8b1 YZ |
445 | u64 oldest_tid; /* oldest incomplete mds request, |
446 | excluding setfilelock requests */ | |
44ca18f2 | 447 | struct rb_root request_tree; /* pending mds requests */ |
2f2dc053 SW |
448 | struct delayed_work delayed_work; /* delayed work */ |
449 | unsigned long last_renew_caps; /* last time we renewed our caps */ | |
450 | struct list_head cap_delay_list; /* caps with delayed release */ | |
451 | spinlock_t cap_delay_lock; /* protects cap_delay_list */ | |
452 | struct list_head snap_flush_list; /* cap_snaps ready to flush */ | |
453 | spinlock_t snap_flush_lock; | |
454 | ||
553adfd9 | 455 | u64 last_cap_flush_tid; |
e4500b5e | 456 | struct list_head cap_flush_list; |
db354052 | 457 | struct list_head cap_dirty_migrating; /* ...that are migration... */ |
2f2dc053 SW |
458 | int num_cap_flushing; /* # caps we are flushing */ |
459 | spinlock_t cap_dirty_lock; /* protects above items */ | |
460 | wait_queue_head_t cap_flushing_wq; | |
461 | ||
37c4efc1 | 462 | struct work_struct cap_reclaim_work; |
fe33032d | 463 | atomic_t cap_reclaim_pending; |
37c4efc1 | 464 | |
37151668 YS |
465 | /* |
466 | * Cap reservations | |
467 | * | |
468 | * Maintain a global pool of preallocated struct ceph_caps, referenced | |
469 | * by struct ceph_caps_reservations. This ensures that we preallocate | |
470 | * memory needed to successfully process an MDS response. (If an MDS | |
471 | * sends us cap information and we fail to process it, we will have | |
472 | * problems due to the client and MDS being out of sync.) | |
473 | * | |
474 | * Reservations are 'owned' by a ceph_cap_reservation context. | |
475 | */ | |
476 | spinlock_t caps_list_lock; | |
477 | struct list_head caps_list; /* unused (reserved or | |
478 | unreserved) */ | |
3a3430af | 479 | struct list_head cap_wait_list; |
37151668 YS |
480 | int caps_total_count; /* total caps allocated */ |
481 | int caps_use_count; /* in use */ | |
fe33032d | 482 | int caps_use_max; /* max used caps */ |
37151668 YS |
483 | int caps_reserve_count; /* unused, reserved */ |
484 | int caps_avail_count; /* unused, unreserved */ | |
485 | int caps_min_count; /* keep at least this many | |
486 | (unreserved) */ | |
37c4efc1 YZ |
487 | spinlock_t dentry_list_lock; |
488 | struct list_head dentry_leases; /* fifo list */ | |
489 | struct list_head dentry_dir_leases; /* lru list */ | |
10183a69 | 490 | |
f9009efa XL |
491 | struct ceph_client_metric metric; |
492 | ||
75c9627e YZ |
493 | spinlock_t snapid_map_lock; |
494 | struct rb_root snapid_map_tree; | |
495 | struct list_head snapid_map_lru; | |
496 | ||
10183a69 YZ |
497 | struct rw_semaphore pool_perm_rwsem; |
498 | struct rb_root pool_perm_tree; | |
717e6f28 YZ |
499 | |
500 | char nodename[__NEW_UTS_LEN + 1]; | |
2f2dc053 SW |
501 | }; |
502 | ||
503 | extern const char *ceph_mds_op_name(int op); | |
504 | ||
3e699bd8 | 505 | extern bool check_session_state(struct ceph_mds_session *s); |
62575e27 | 506 | void inc_session_sequence(struct ceph_mds_session *s); |
3e699bd8 | 507 | |
2f2dc053 SW |
508 | extern struct ceph_mds_session * |
509 | __ceph_lookup_mds_session(struct ceph_mds_client *, int mds); | |
510 | ||
a687ecaf JS |
511 | extern const char *ceph_session_state_name(int s); |
512 | ||
5b3248c6 XL |
513 | extern struct ceph_mds_session * |
514 | ceph_get_mds_session(struct ceph_mds_session *s); | |
2f2dc053 SW |
515 | extern void ceph_put_mds_session(struct ceph_mds_session *s); |
516 | ||
517 | extern int ceph_send_msg_mds(struct ceph_mds_client *mdsc, | |
518 | struct ceph_msg *msg, int mds); | |
519 | ||
3d14c5d2 | 520 | extern int ceph_mdsc_init(struct ceph_fs_client *fsc); |
2f2dc053 | 521 | extern void ceph_mdsc_close_sessions(struct ceph_mds_client *mdsc); |
48fec5d0 | 522 | extern void ceph_mdsc_force_umount(struct ceph_mds_client *mdsc); |
3d14c5d2 | 523 | extern void ceph_mdsc_destroy(struct ceph_fs_client *fsc); |
2f2dc053 SW |
524 | |
525 | extern void ceph_mdsc_sync(struct ceph_mds_client *mdsc); | |
526 | ||
167c9e35 | 527 | extern void ceph_invalidate_dir_request(struct ceph_mds_request *req); |
54008399 YZ |
528 | extern int ceph_alloc_readdir_reply_buffer(struct ceph_mds_request *req, |
529 | struct inode *dir); | |
2f2dc053 SW |
530 | extern struct ceph_mds_request * |
531 | ceph_mdsc_create_request(struct ceph_mds_client *mdsc, int op, int mode); | |
86bda539 JL |
532 | extern int ceph_mdsc_submit_request(struct ceph_mds_client *mdsc, |
533 | struct inode *dir, | |
534 | struct ceph_mds_request *req); | |
9eaa7b79 JL |
535 | int ceph_mdsc_wait_request(struct ceph_mds_client *mdsc, |
536 | struct ceph_mds_request *req, | |
537 | ceph_mds_request_wait_callback_t wait_func); | |
2f2dc053 SW |
538 | extern int ceph_mdsc_do_request(struct ceph_mds_client *mdsc, |
539 | struct inode *dir, | |
540 | struct ceph_mds_request *req); | |
a25949b9 | 541 | extern void ceph_mdsc_release_dir_caps(struct ceph_mds_request *req); |
e64f44a8 | 542 | extern void ceph_mdsc_release_dir_caps_no_check(struct ceph_mds_request *req); |
2f2dc053 SW |
543 | static inline void ceph_mdsc_get_request(struct ceph_mds_request *req) |
544 | { | |
153c8e6b SW |
545 | kref_get(&req->r_kref); |
546 | } | |
547 | extern void ceph_mdsc_release_request(struct kref *kref); | |
548 | static inline void ceph_mdsc_put_request(struct ceph_mds_request *req) | |
549 | { | |
550 | kref_put(&req->r_kref, ceph_mdsc_release_request); | |
2f2dc053 | 551 | } |
2f2dc053 | 552 | |
d095559c | 553 | extern void send_flush_mdlog(struct ceph_mds_session *s); |
59b312f3 XL |
554 | extern void ceph_mdsc_iterate_sessions(struct ceph_mds_client *mdsc, |
555 | void (*cb)(struct ceph_mds_session *), | |
556 | bool check_state); | |
fba97e80 | 557 | extern struct ceph_msg *ceph_create_session_msg(u32 op, u64 seq); |
e3ec8d68 YZ |
558 | extern void __ceph_queue_cap_release(struct ceph_mds_session *session, |
559 | struct ceph_cap *cap); | |
560 | extern void ceph_flush_cap_releases(struct ceph_mds_client *mdsc, | |
561 | struct ceph_mds_session *session); | |
37c4efc1 | 562 | extern void ceph_queue_cap_reclaim_work(struct ceph_mds_client *mdsc); |
fe33032d | 563 | extern void ceph_reclaim_caps_nr(struct ceph_mds_client *mdsc, int nr); |
f5d77269 | 564 | extern int ceph_iterate_session_caps(struct ceph_mds_session *session, |
aaf67de7 | 565 | int (*cb)(struct inode *, int mds, void *), |
f5d77269 | 566 | void *arg); |
2f2dc053 SW |
567 | extern void ceph_mdsc_pre_umount(struct ceph_mds_client *mdsc); |
568 | ||
f77f21bb JL |
569 | static inline void ceph_mdsc_free_path(char *path, int len) |
570 | { | |
2a575f13 | 571 | if (!IS_ERR_OR_NULL(path)) |
f77f21bb JL |
572 | __putname(path - (PATH_MAX - 1 - len)); |
573 | } | |
574 | ||
2f2dc053 | 575 | extern char *ceph_mdsc_build_path(struct dentry *dentry, int *plen, u64 *base, |
3fd945a7 | 576 | int for_wire); |
2f2dc053 SW |
577 | |
578 | extern void __ceph_mdsc_drop_dentry_lease(struct dentry *dentry); | |
579 | extern void ceph_mdsc_lease_send_msg(struct ceph_mds_session *session, | |
2f2dc053 SW |
580 | struct dentry *dentry, char action, |
581 | u32 seq); | |
582 | ||
430afbad YZ |
583 | extern void ceph_mdsc_handle_mdsmap(struct ceph_mds_client *mdsc, |
584 | struct ceph_msg *msg); | |
585 | extern void ceph_mdsc_handle_fsmap(struct ceph_mds_client *mdsc, | |
586 | struct ceph_msg *msg); | |
2f2dc053 | 587 | |
5d72d13c YZ |
588 | extern struct ceph_mds_session * |
589 | ceph_mdsc_open_export_target_session(struct ceph_mds_client *mdsc, int target); | |
154f42c2 SW |
590 | extern void ceph_mdsc_open_export_target_sessions(struct ceph_mds_client *mdsc, |
591 | struct ceph_mds_session *session); | |
592 | ||
e30ee581 ZZ |
593 | extern int ceph_trim_caps(struct ceph_mds_client *mdsc, |
594 | struct ceph_mds_session *session, | |
595 | int max_caps); | |
d4846487 | 596 | |
891f3f5a JL |
597 | static inline int ceph_wait_on_async_create(struct inode *inode) |
598 | { | |
599 | struct ceph_inode_info *ci = ceph_inode(inode); | |
600 | ||
601 | return wait_on_bit(&ci->i_ceph_flags, CEPH_ASYNC_CREATE_BIT, | |
5e56776d | 602 | TASK_KILLABLE); |
891f3f5a | 603 | } |
d4846487 | 604 | |
4868e537 | 605 | extern int ceph_wait_on_conflict_unlink(struct dentry *dentry); |
d4846487 JL |
606 | extern u64 ceph_get_deleg_ino(struct ceph_mds_session *session); |
607 | extern int ceph_restore_deleg_ino(struct ceph_mds_session *session, u64 ino); | |
2f2dc053 | 608 | #endif |