1 | /* SPDX-License-Identifier: GPL-2.0 */ |
2 | #ifndef _FS_CEPH_SUPER_H |
3 | #define _FS_CEPH_SUPER_H |
4 | |
5 | #include <linux/ceph/ceph_debug.h> |
6 | #include <linux/ceph/osd_client.h> |
7 | |
8 | #include <asm/unaligned.h> |
9 | #include <linux/backing-dev.h> |
10 | #include <linux/completion.h> |
11 | #include <linux/exportfs.h> |
12 | #include <linux/fs.h> |
13 | #include <linux/mempool.h> |
14 | #include <linux/pagemap.h> |
15 | #include <linux/wait.h> |
16 | #include <linux/writeback.h> |
17 | #include <linux/slab.h> |
18 | #include <linux/posix_acl.h> |
19 | #include <linux/refcount.h> |
20 | #include <linux/security.h> |
21 | #include <linux/netfs.h> |
22 | #include <linux/fscache.h> |
23 | #include <linux/hashtable.h> |
24 | |
25 | #include <linux/ceph/libceph.h> |
26 | #include "crypto.h" |
27 | |
28 | /* large granularity for statfs utilization stats to facilitate |
29 | * large volume sizes on 32-bit machines. */ |
30 | #define CEPH_BLOCK_SHIFT 22 /* 4 MB */ |
31 | #define CEPH_BLOCK (1 << CEPH_BLOCK_SHIFT) |
32 | #define CEPH_4K_BLOCK_SHIFT 12 /* 4 KB */ |
33 | |
34 | #define CEPH_MOUNT_OPT_CLEANRECOVER (1<<1) /* auto reonnect (clean mode) after blocklisted */ |
35 | #define CEPH_MOUNT_OPT_DIRSTAT (1<<4) /* `cat dirname` for stats */ |
36 | #define CEPH_MOUNT_OPT_RBYTES (1<<5) /* dir st_bytes = rbytes */ |
37 | #define CEPH_MOUNT_OPT_NOASYNCREADDIR (1<<7) /* no dcache readdir */ |
38 | #define CEPH_MOUNT_OPT_INO32 (1<<8) /* 32 bit inos */ |
39 | #define CEPH_MOUNT_OPT_DCACHE (1<<9) /* use dcache for readdir etc */ |
40 | #define CEPH_MOUNT_OPT_FSCACHE (1<<10) /* use fscache */ |
41 | #define CEPH_MOUNT_OPT_NOPOOLPERM (1<<11) /* no pool permission check */ |
42 | #define CEPH_MOUNT_OPT_MOUNTWAIT (1<<12) /* mount waits if no mds is up */ |
43 | #define CEPH_MOUNT_OPT_NOQUOTADF (1<<13) /* no root dir quota in statfs */ |
44 | #define CEPH_MOUNT_OPT_NOCOPYFROM (1<<14) /* don't use RADOS 'copy-from' op */ |
45 | #define CEPH_MOUNT_OPT_ASYNC_DIROPS (1<<15) /* allow async directory ops */ |
46 | #define CEPH_MOUNT_OPT_NOPAGECACHE (1<<16) /* bypass pagecache altogether */ |
47 | #define CEPH_MOUNT_OPT_SPARSEREAD (1<<17) /* always do sparse reads */ |
48 | |
49 | #define CEPH_MOUNT_OPT_DEFAULT \ |
50 | (CEPH_MOUNT_OPT_DCACHE | \ |
51 | CEPH_MOUNT_OPT_NOCOPYFROM | \ |
52 | CEPH_MOUNT_OPT_ASYNC_DIROPS) |
53 | |
54 | #define ceph_set_mount_opt(fsc, opt) \ |
55 | (fsc)->mount_options->flags |= CEPH_MOUNT_OPT_##opt |
56 | #define ceph_clear_mount_opt(fsc, opt) \ |
57 | (fsc)->mount_options->flags &= ~CEPH_MOUNT_OPT_##opt |
58 | #define ceph_test_mount_opt(fsc, opt) \ |
59 | (!!((fsc)->mount_options->flags & CEPH_MOUNT_OPT_##opt)) |
60 | |
61 | /* max size of osd read request, limited by libceph */ |
62 | #define CEPH_MAX_READ_SIZE CEPH_MSG_MAX_DATA_LEN |
63 | /* osd has a configurable limitaion of max write size. |
64 | * CEPH_MSG_MAX_DATA_LEN should be small enough. */ |
65 | #define CEPH_MAX_WRITE_SIZE CEPH_MSG_MAX_DATA_LEN |
66 | #define CEPH_RASIZE_DEFAULT (8192*1024) /* max readahead */ |
67 | #define CEPH_MAX_READDIR_DEFAULT 1024 |
68 | #define CEPH_MAX_READDIR_BYTES_DEFAULT (512*1024) |
69 | #define CEPH_SNAPDIRNAME_DEFAULT ".snap" |
70 | |
71 | /* |
72 | * Delay telling the MDS we no longer want caps, in case we reopen |
73 | * the file. Delay a minimum amount of time, even if we send a cap |
74 | * message for some other reason. Otherwise, take the oppotunity to |
75 | * update the mds to avoid sending another message later. |
76 | */ |
77 | #define CEPH_CAPS_WANTED_DELAY_MIN_DEFAULT 5 /* cap release delay */ |
78 | #define CEPH_CAPS_WANTED_DELAY_MAX_DEFAULT 60 /* cap release delay */ |
79 | |
80 | struct ceph_mount_options { |
81 | unsigned int flags; |
82 | |
83 | unsigned int wsize; /* max write size */ |
84 | unsigned int rsize; /* max read size */ |
85 | unsigned int rasize; /* max readahead */ |
86 | unsigned int congestion_kb; /* max writeback in flight */ |
87 | unsigned int caps_wanted_delay_min, caps_wanted_delay_max; |
88 | int caps_max; |
89 | unsigned int max_readdir; /* max readdir result (entries) */ |
90 | unsigned int max_readdir_bytes; /* max readdir result (bytes) */ |
91 | |
92 | bool new_dev_syntax; |
93 | |
94 | /* |
95 | * everything above this point can be memcmp'd; everything below |
96 | * is handled in compare_mount_options() |
97 | */ |
98 | |
99 | char *snapdir_name; /* default ".snap" */ |
100 | char *mds_namespace; /* default NULL */ |
101 | char *server_path; /* default NULL (means "/") */ |
102 | char *fscache_uniq; /* default NULL */ |
103 | char *mon_addr; |
104 | struct fscrypt_dummy_policy dummy_enc_policy; |
105 | }; |
106 | |
107 | /* mount state */ |
108 | enum { |
109 | CEPH_MOUNT_MOUNTING, |
110 | CEPH_MOUNT_MOUNTED, |
111 | CEPH_MOUNT_UNMOUNTING, |
112 | CEPH_MOUNT_UNMOUNTED, |
113 | CEPH_MOUNT_SHUTDOWN, |
114 | CEPH_MOUNT_RECOVER, |
115 | CEPH_MOUNT_FENCE_IO, |
116 | }; |
117 | |
118 | #define CEPH_ASYNC_CREATE_CONFLICT_BITS 8 |
119 | |
120 | struct ceph_fs_client { |
121 | struct super_block *sb; |
122 | |
123 | struct list_head metric_wakeup; |
124 | |
125 | struct ceph_mount_options *mount_options; |
126 | struct ceph_client *client; |
127 | |
128 | int mount_state; |
129 | |
130 | bool blocklisted; |
131 | |
132 | bool have_copy_from2; |
133 | |
134 | u32 filp_gen; |
135 | loff_t max_file_size; |
136 | |
137 | struct ceph_mds_client *mdsc; |
138 | |
139 | atomic_long_t writeback_count; |
140 | bool write_congested; |
141 | |
142 | struct workqueue_struct *inode_wq; |
143 | struct workqueue_struct *cap_wq; |
144 | |
145 | DECLARE_HASHTABLE(async_unlink_conflict, CEPH_ASYNC_CREATE_CONFLICT_BITS); |
146 | spinlock_t async_unlink_conflict_lock; |
147 | |
148 | #ifdef CONFIG_DEBUG_FS |
149 | struct dentry *debugfs_dentry_lru, *debugfs_caps; |
150 | struct dentry *debugfs_congestion_kb; |
151 | struct dentry *debugfs_bdi; |
152 | struct dentry *debugfs_mdsc, *debugfs_mdsmap; |
153 | struct dentry *debugfs_status; |
154 | struct dentry *debugfs_mds_sessions; |
155 | struct dentry *debugfs_metrics_dir; |
156 | #endif |
157 | |
158 | #ifdef CONFIG_CEPH_FSCACHE |
159 | struct fscache_volume *fscache; |
160 | #endif |
161 | #ifdef CONFIG_FS_ENCRYPTION |
162 | struct fscrypt_dummy_policy fsc_dummy_enc_policy; |
163 | #endif |
164 | }; |
165 | |
166 | /* |
167 | * File i/o capability. This tracks shared state with the metadata |
168 | * server that allows us to cache or writeback attributes or to read |
169 | * and write data. For any given inode, we should have one or more |
170 | * capabilities, one issued by each metadata server, and our |
171 | * cumulative access is the OR of all issued capabilities. |
172 | * |
173 | * Each cap is referenced by the inode's i_caps rbtree and by per-mds |
174 | * session capability lists. |
175 | */ |
176 | struct ceph_cap { |
177 | struct ceph_inode_info *ci; |
178 | struct rb_node ci_node; /* per-ci cap tree */ |
179 | struct ceph_mds_session *session; |
180 | struct list_head session_caps; /* per-session caplist */ |
181 | u64 cap_id; /* unique cap id (mds provided) */ |
182 | union { |
183 | /* in-use caps */ |
184 | struct { |
185 | int issued; /* latest, from the mds */ |
186 | int implemented; /* implemented superset of |
187 | issued (for revocation) */ |
188 | int mds; /* mds index for this cap */ |
189 | int mds_wanted; /* caps wanted from this mds */ |
190 | }; |
191 | /* caps to release */ |
192 | struct { |
193 | u64 cap_ino; |
194 | int queue_release; |
195 | }; |
196 | }; |
197 | u32 seq, issue_seq, mseq; |
198 | u32 cap_gen; /* active/stale cycle */ |
199 | unsigned long last_used; |
200 | struct list_head caps_item; |
201 | }; |
202 | |
203 | #define CHECK_CAPS_AUTHONLY 1 /* only check auth cap */ |
204 | #define CHECK_CAPS_FLUSH 2 /* flush any dirty caps */ |
205 | #define CHECK_CAPS_NOINVAL 4 /* don't invalidate pagecache */ |
206 | |
207 | struct ceph_cap_flush { |
208 | u64 tid; |
209 | int caps; |
210 | bool wake; /* wake up flush waiters when finish ? */ |
211 | bool is_capsnap; /* true means capsnap */ |
212 | struct list_head g_list; // global |
213 | struct list_head i_list; // per inode |
214 | }; |
215 | |
216 | /* |
217 | * Snapped cap state that is pending flush to mds. When a snapshot occurs, |
218 | * we first complete any in-process sync writes and writeback any dirty |
219 | * data before flushing the snapped state (tracked here) back to the MDS. |
220 | */ |
221 | struct ceph_cap_snap { |
222 | refcount_t nref; |
223 | struct list_head ci_item; |
224 | |
225 | struct ceph_cap_flush cap_flush; |
226 | |
227 | u64 follows; |
228 | int issued, dirty; |
229 | struct ceph_snap_context *context; |
230 | |
231 | umode_t mode; |
232 | kuid_t uid; |
233 | kgid_t gid; |
234 | |
235 | struct ceph_buffer *xattr_blob; |
236 | u64 xattr_version; |
237 | |
238 | u64 size; |
239 | u64 change_attr; |
240 | struct timespec64 mtime, atime, ctime, btime; |
241 | u64 time_warp_seq; |
242 | u64 truncate_size; |
243 | u32 truncate_seq; |
244 | int writing; /* a sync write is still in progress */ |
245 | int dirty_pages; /* dirty pages awaiting writeback */ |
246 | bool inline_data; |
247 | bool need_flush; |
248 | }; |
249 | |
250 | static inline void ceph_put_cap_snap(struct ceph_cap_snap *capsnap) |
251 | { |
252 | if (refcount_dec_and_test(r: &capsnap->nref)) { |
253 | if (capsnap->xattr_blob) |
254 | ceph_buffer_put(b: capsnap->xattr_blob); |
255 | kmem_cache_free(s: ceph_cap_snap_cachep, objp: capsnap); |
256 | } |
257 | } |
258 | |
259 | /* |
260 | * The frag tree describes how a directory is fragmented, potentially across |
261 | * multiple metadata servers. It is also used to indicate points where |
262 | * metadata authority is delegated, and whether/where metadata is replicated. |
263 | * |
264 | * A _leaf_ frag will be present in the i_fragtree IFF there is |
265 | * delegation info. That is, if mds >= 0 || ndist > 0. |
266 | */ |
267 | #define CEPH_MAX_DIRFRAG_REP 4 |
268 | |
269 | struct ceph_inode_frag { |
270 | struct rb_node node; |
271 | |
272 | /* fragtree state */ |
273 | u32 frag; |
274 | int split_by; /* i.e. 2^(split_by) children */ |
275 | |
276 | /* delegation and replication info */ |
277 | int mds; /* -1 if same authority as parent */ |
278 | int ndist; /* >0 if replicated */ |
279 | int dist[CEPH_MAX_DIRFRAG_REP]; |
280 | }; |
281 | |
282 | /* |
283 | * We cache inode xattrs as an encoded blob until they are first used, |
284 | * at which point we parse them into an rbtree. |
285 | */ |
286 | struct ceph_inode_xattr { |
287 | struct rb_node node; |
288 | |
289 | const char *name; |
290 | int name_len; |
291 | const char *val; |
292 | int val_len; |
293 | int dirty; |
294 | |
295 | int should_free_name; |
296 | int should_free_val; |
297 | }; |
298 | |
299 | /* |
300 | * Ceph dentry state |
301 | */ |
302 | struct ceph_dentry_info { |
303 | struct dentry *dentry; |
304 | struct ceph_mds_session *lease_session; |
305 | struct list_head lease_list; |
306 | struct hlist_node hnode; |
307 | unsigned long flags; |
308 | int lease_shared_gen; |
309 | u32 lease_gen; |
310 | u32 lease_seq; |
311 | unsigned long lease_renew_after, lease_renew_from; |
312 | unsigned long time; |
313 | u64 offset; |
314 | }; |
315 | |
316 | #define CEPH_DENTRY_REFERENCED (1 << 0) |
317 | #define CEPH_DENTRY_LEASE_LIST (1 << 1) |
318 | #define CEPH_DENTRY_SHRINK_LIST (1 << 2) |
319 | #define CEPH_DENTRY_PRIMARY_LINK (1 << 3) |
320 | #define CEPH_DENTRY_ASYNC_UNLINK_BIT (4) |
321 | #define CEPH_DENTRY_ASYNC_UNLINK (1 << CEPH_DENTRY_ASYNC_UNLINK_BIT) |
322 | #define CEPH_DENTRY_ASYNC_CREATE_BIT (5) |
323 | #define CEPH_DENTRY_ASYNC_CREATE (1 << CEPH_DENTRY_ASYNC_CREATE_BIT) |
324 | |
325 | struct ceph_inode_xattrs_info { |
326 | /* |
327 | * (still encoded) xattr blob. we avoid the overhead of parsing |
328 | * this until someone actually calls getxattr, etc. |
329 | * |
330 | * blob->vec.iov_len == 4 implies there are no xattrs; blob == |
331 | * NULL means we don't know. |
332 | */ |
333 | struct ceph_buffer *blob, *prealloc_blob; |
334 | |
335 | struct rb_root index; |
336 | bool dirty; |
337 | int count; |
338 | int names_size; |
339 | int vals_size; |
340 | u64 version, index_version; |
341 | }; |
342 | |
343 | /* |
344 | * Ceph inode. |
345 | */ |
346 | struct ceph_inode_info { |
347 | struct netfs_inode netfs; /* Netfslib context and vfs inode */ |
348 | struct ceph_vino i_vino; /* ceph ino + snap */ |
349 | |
350 | spinlock_t i_ceph_lock; |
351 | |
352 | u64 i_version; |
353 | u64 i_inline_version; |
354 | u32 i_time_warp_seq; |
355 | |
356 | unsigned long i_ceph_flags; |
357 | atomic64_t i_release_count; |
358 | atomic64_t i_ordered_count; |
359 | atomic64_t i_complete_seq[2]; |
360 | |
361 | struct ceph_dir_layout i_dir_layout; |
362 | struct ceph_file_layout i_layout; |
363 | struct ceph_file_layout i_cached_layout; // for async creates |
364 | char *i_symlink; |
365 | |
366 | /* for dirs */ |
367 | struct timespec64 i_rctime; |
368 | u64 i_rbytes, i_rfiles, i_rsubdirs, i_rsnaps; |
369 | u64 i_files, i_subdirs; |
370 | |
371 | /* quotas */ |
372 | u64 i_max_bytes, i_max_files; |
373 | |
374 | s32 i_dir_pin; |
375 | |
376 | struct rb_root i_fragtree; |
377 | int i_fragtree_nsplits; |
378 | struct mutex i_fragtree_mutex; |
379 | |
380 | struct ceph_inode_xattrs_info i_xattrs; |
381 | |
382 | /* capabilities. protected _both_ by i_ceph_lock and cap->session's |
383 | * s_mutex. */ |
384 | struct rb_root i_caps; /* cap list */ |
385 | struct ceph_cap *i_auth_cap; /* authoritative cap, if any */ |
386 | unsigned i_dirty_caps, i_flushing_caps; /* mask of dirtied fields */ |
387 | |
388 | /* |
389 | * Link to the auth cap's session's s_cap_dirty list. s_cap_dirty |
390 | * is protected by the mdsc->cap_dirty_lock, but each individual item |
391 | * is also protected by the inode's i_ceph_lock. Walking s_cap_dirty |
392 | * requires the mdsc->cap_dirty_lock. List presence for an item can |
393 | * be tested under the i_ceph_lock. Changing anything requires both. |
394 | */ |
395 | struct list_head i_dirty_item; |
396 | |
397 | /* |
398 | * Link to session's s_cap_flushing list. Protected in a similar |
399 | * fashion to i_dirty_item, but also by the s_mutex for changes. The |
400 | * s_cap_flushing list can be walked while holding either the s_mutex |
401 | * or msdc->cap_dirty_lock. List presence can also be checked while |
402 | * holding the i_ceph_lock for this inode. |
403 | */ |
404 | struct list_head i_flushing_item; |
405 | |
406 | /* we need to track cap writeback on a per-cap-bit basis, to allow |
407 | * overlapping, pipelined cap flushes to the mds. we can probably |
408 | * reduce the tid to 8 bits if we're concerned about inode size. */ |
409 | struct ceph_cap_flush *i_prealloc_cap_flush; |
410 | struct list_head i_cap_flush_list; |
411 | wait_queue_head_t i_cap_wq; /* threads waiting on a capability */ |
412 | unsigned long i_hold_caps_max; /* jiffies */ |
413 | struct list_head i_cap_delay_list; /* for delayed cap release to mds */ |
414 | struct ceph_cap_reservation i_cap_migration_resv; |
415 | struct list_head i_cap_snaps; /* snapped state pending flush to mds */ |
416 | struct ceph_snap_context *i_head_snapc; /* set if wr_buffer_head > 0 or |
417 | dirty|flushing caps */ |
418 | unsigned i_snap_caps; /* cap bits for snapped files */ |
419 | |
420 | unsigned long i_last_rd; |
421 | unsigned long i_last_wr; |
422 | int i_nr_by_mode[CEPH_FILE_MODE_BITS]; /* open file counts */ |
423 | |
424 | struct mutex i_truncate_mutex; |
425 | u32 i_truncate_seq; /* last truncate to smaller size */ |
426 | u64 i_truncate_size; /* and the size we last truncated down to */ |
427 | int i_truncate_pending; /* still need to call vmtruncate */ |
428 | /* |
429 | * For none fscrypt case it equals to i_truncate_size or it will |
430 | * equals to fscrypt_file_size |
431 | */ |
432 | u64 i_truncate_pagecache_size; |
433 | |
434 | u64 i_max_size; /* max file size authorized by mds */ |
435 | u64 i_reported_size; /* (max_)size reported to or requested of mds */ |
436 | u64 i_wanted_max_size; /* offset we'd like to write too */ |
437 | u64 i_requested_max_size; /* max_size we've requested */ |
438 | |
439 | /* held references to caps */ |
440 | int i_pin_ref; |
441 | int i_rd_ref, i_rdcache_ref, i_wr_ref, i_wb_ref, i_fx_ref; |
442 | int i_wrbuffer_ref, i_wrbuffer_ref_head; |
443 | atomic_t i_filelock_ref; |
444 | atomic_t i_shared_gen; /* increment each time we get FILE_SHARED */ |
445 | u32 i_rdcache_gen; /* incremented each time we get FILE_CACHE. */ |
446 | u32 i_rdcache_revoking; /* RDCACHE gen to async invalidate, if any */ |
447 | |
448 | struct list_head i_unsafe_dirops; /* uncommitted mds dir ops */ |
449 | struct list_head i_unsafe_iops; /* uncommitted mds inode ops */ |
450 | spinlock_t i_unsafe_lock; |
451 | |
452 | union { |
453 | struct ceph_snap_realm *i_snap_realm; /* snap realm (if caps) */ |
454 | struct ceph_snapid_map *i_snapid_map; /* snapid -> dev_t */ |
455 | }; |
456 | struct list_head i_snap_realm_item; |
457 | struct list_head i_snap_flush_item; |
458 | struct timespec64 i_btime; |
459 | struct timespec64 i_snap_btime; |
460 | |
461 | struct work_struct i_work; |
462 | unsigned long i_work_mask; |
463 | |
464 | #ifdef CONFIG_FS_ENCRYPTION |
465 | u32 fscrypt_auth_len; |
466 | u32 fscrypt_file_len; |
467 | u8 *fscrypt_auth; |
468 | u8 *fscrypt_file; |
469 | #endif |
470 | }; |
471 | |
472 | struct ceph_netfs_request_data { |
473 | int caps; |
474 | |
475 | /* |
476 | * Maximum size of a file readahead request. |
477 | * The fadvise could update the bdi's default ra_pages. |
478 | */ |
479 | unsigned int file_ra_pages; |
480 | |
481 | /* Set it if fadvise disables file readahead entirely */ |
482 | bool file_ra_disabled; |
483 | }; |
484 | |
485 | static inline struct ceph_inode_info * |
486 | ceph_inode(const struct inode *inode) |
487 | { |
488 | return container_of(inode, struct ceph_inode_info, netfs.inode); |
489 | } |
490 | |
491 | static inline struct ceph_fs_client * |
492 | ceph_inode_to_fs_client(const struct inode *inode) |
493 | { |
494 | return (struct ceph_fs_client *)inode->i_sb->s_fs_info; |
495 | } |
496 | |
497 | static inline struct ceph_fs_client * |
498 | ceph_sb_to_fs_client(const struct super_block *sb) |
499 | { |
500 | return (struct ceph_fs_client *)sb->s_fs_info; |
501 | } |
502 | |
503 | static inline struct ceph_mds_client * |
504 | ceph_sb_to_mdsc(const struct super_block *sb) |
505 | { |
506 | return (struct ceph_mds_client *)ceph_sb_to_fs_client(sb)->mdsc; |
507 | } |
508 | |
509 | static inline struct ceph_client * |
510 | ceph_inode_to_client(const struct inode *inode) |
511 | { |
512 | return (struct ceph_client *)ceph_inode_to_fs_client(inode)->client; |
513 | } |
514 | |
515 | static inline struct ceph_vino |
516 | ceph_vino(const struct inode *inode) |
517 | { |
518 | return ceph_inode(inode)->i_vino; |
519 | } |
520 | |
521 | static inline u32 ceph_ino_to_ino32(u64 vino) |
522 | { |
523 | u32 ino = vino & 0xffffffff; |
524 | ino ^= vino >> 32; |
525 | if (!ino) |
526 | ino = 2; |
527 | return ino; |
528 | } |
529 | |
530 | /* |
531 | * Inode numbers in cephfs are 64 bits, but inode->i_ino is 32-bits on |
532 | * some arches. We generally do not use this value inside the ceph driver, but |
533 | * we do want to set it to something, so that generic vfs code has an |
534 | * appropriate value for tracepoints and the like. |
535 | */ |
536 | static inline ino_t ceph_vino_to_ino_t(struct ceph_vino vino) |
537 | { |
538 | if (sizeof(ino_t) == sizeof(u32)) |
539 | return ceph_ino_to_ino32(vino: vino.ino); |
540 | return (ino_t)vino.ino; |
541 | } |
542 | |
543 | /* for printf-style formatting */ |
544 | #define ceph_vinop(i) ceph_inode(i)->i_vino.ino, ceph_inode(i)->i_vino.snap |
545 | |
546 | static inline u64 ceph_ino(struct inode *inode) |
547 | { |
548 | return ceph_inode(inode)->i_vino.ino; |
549 | } |
550 | |
551 | static inline u64 ceph_snap(struct inode *inode) |
552 | { |
553 | return ceph_inode(inode)->i_vino.snap; |
554 | } |
555 | |
556 | /** |
557 | * ceph_present_ino - format an inode number for presentation to userland |
558 | * @sb: superblock where the inode lives |
559 | * @ino: inode number to (possibly) convert |
560 | * |
561 | * If the user mounted with the ino32 option, then the 64-bit value needs |
562 | * to be converted to something that can fit inside 32 bits. Note that |
563 | * internal kernel code never uses this value, so this is entirely for |
564 | * userland consumption. |
565 | */ |
566 | static inline u64 ceph_present_ino(struct super_block *sb, u64 ino) |
567 | { |
568 | if (unlikely(ceph_test_mount_opt(ceph_sb_to_fs_client(sb), INO32))) |
569 | return ceph_ino_to_ino32(vino: ino); |
570 | return ino; |
571 | } |
572 | |
573 | static inline u64 ceph_present_inode(struct inode *inode) |
574 | { |
575 | return ceph_present_ino(sb: inode->i_sb, ino: ceph_ino(inode)); |
576 | } |
577 | |
578 | static inline int ceph_ino_compare(struct inode *inode, void *data) |
579 | { |
580 | struct ceph_vino *pvino = (struct ceph_vino *)data; |
581 | struct ceph_inode_info *ci = ceph_inode(inode); |
582 | return ci->i_vino.ino == pvino->ino && |
583 | ci->i_vino.snap == pvino->snap; |
584 | } |
585 | |
586 | /* |
587 | * The MDS reserves a set of inodes for its own usage. These should never |
588 | * be accessible by clients, and so the MDS has no reason to ever hand these |
589 | * out. The range is CEPH_MDS_INO_MDSDIR_OFFSET..CEPH_INO_SYSTEM_BASE. |
590 | * |
591 | * These come from src/mds/mdstypes.h in the ceph sources. |
592 | */ |
593 | #define CEPH_MAX_MDS 0x100 |
594 | #define CEPH_NUM_STRAY 10 |
595 | #define CEPH_MDS_INO_MDSDIR_OFFSET (1 * CEPH_MAX_MDS) |
596 | #define CEPH_MDS_INO_LOG_OFFSET (2 * CEPH_MAX_MDS) |
597 | #define CEPH_INO_SYSTEM_BASE ((6*CEPH_MAX_MDS) + (CEPH_MAX_MDS * CEPH_NUM_STRAY)) |
598 | |
599 | static inline bool ceph_vino_is_reserved(const struct ceph_vino vino) |
600 | { |
601 | if (vino.ino >= CEPH_INO_SYSTEM_BASE || |
602 | vino.ino < CEPH_MDS_INO_MDSDIR_OFFSET) |
603 | return false; |
604 | |
605 | /* Don't warn on mdsdirs */ |
606 | WARN_RATELIMIT(vino.ino >= CEPH_MDS_INO_LOG_OFFSET, |
607 | "Attempt to access reserved inode number 0x%llx" , |
608 | vino.ino); |
609 | return true; |
610 | } |
611 | |
612 | static inline struct inode *ceph_find_inode(struct super_block *sb, |
613 | struct ceph_vino vino) |
614 | { |
615 | if (ceph_vino_is_reserved(vino)) |
616 | return NULL; |
617 | |
618 | /* |
619 | * NB: The hashval will be run through the fs/inode.c hash function |
620 | * anyway, so there is no need to squash the inode number down to |
621 | * 32-bits first. Just use low-order bits on arches with 32-bit long. |
622 | */ |
623 | return ilookup5(sb, hashval: (unsigned long)vino.ino, test: ceph_ino_compare, data: &vino); |
624 | } |
625 | |
626 | |
627 | /* |
628 | * Ceph inode. |
629 | */ |
630 | #define CEPH_I_DIR_ORDERED (1 << 0) /* dentries in dir are ordered */ |
631 | #define CEPH_I_FLUSH (1 << 2) /* do not delay flush of dirty metadata */ |
632 | #define CEPH_I_POOL_PERM (1 << 3) /* pool rd/wr bits are valid */ |
633 | #define CEPH_I_POOL_RD (1 << 4) /* can read from pool */ |
634 | #define CEPH_I_POOL_WR (1 << 5) /* can write to pool */ |
635 | #define CEPH_I_SEC_INITED (1 << 6) /* security initialized */ |
636 | #define CEPH_I_KICK_FLUSH (1 << 7) /* kick flushing caps */ |
637 | #define CEPH_I_FLUSH_SNAPS (1 << 8) /* need flush snapss */ |
638 | #define CEPH_I_ERROR_WRITE (1 << 9) /* have seen write errors */ |
639 | #define CEPH_I_ERROR_FILELOCK (1 << 10) /* have seen file lock errors */ |
640 | #define CEPH_I_ODIRECT (1 << 11) /* inode in direct I/O mode */ |
641 | #define CEPH_ASYNC_CREATE_BIT (12) /* async create in flight for this */ |
642 | #define CEPH_I_ASYNC_CREATE (1 << CEPH_ASYNC_CREATE_BIT) |
643 | #define CEPH_I_SHUTDOWN (1 << 13) /* inode is no longer usable */ |
644 | #define CEPH_I_ASYNC_CHECK_CAPS (1 << 14) /* check caps immediately after async |
645 | creating finishes */ |
646 | |
647 | /* |
648 | * Masks of ceph inode work. |
649 | */ |
650 | #define CEPH_I_WORK_WRITEBACK 0 |
651 | #define CEPH_I_WORK_INVALIDATE_PAGES 1 |
652 | #define CEPH_I_WORK_VMTRUNCATE 2 |
653 | #define CEPH_I_WORK_CHECK_CAPS 3 |
654 | #define CEPH_I_WORK_FLUSH_SNAPS 4 |
655 | |
656 | /* |
657 | * We set the ERROR_WRITE bit when we start seeing write errors on an inode |
658 | * and then clear it when they start succeeding. Note that we do a lockless |
659 | * check first, and only take the lock if it looks like it needs to be changed. |
660 | * The write submission code just takes this as a hint, so we're not too |
661 | * worried if a few slip through in either direction. |
662 | */ |
663 | static inline void ceph_set_error_write(struct ceph_inode_info *ci) |
664 | { |
665 | if (!(READ_ONCE(ci->i_ceph_flags) & CEPH_I_ERROR_WRITE)) { |
666 | spin_lock(lock: &ci->i_ceph_lock); |
667 | ci->i_ceph_flags |= CEPH_I_ERROR_WRITE; |
668 | spin_unlock(lock: &ci->i_ceph_lock); |
669 | } |
670 | } |
671 | |
672 | static inline void ceph_clear_error_write(struct ceph_inode_info *ci) |
673 | { |
674 | if (READ_ONCE(ci->i_ceph_flags) & CEPH_I_ERROR_WRITE) { |
675 | spin_lock(lock: &ci->i_ceph_lock); |
676 | ci->i_ceph_flags &= ~CEPH_I_ERROR_WRITE; |
677 | spin_unlock(lock: &ci->i_ceph_lock); |
678 | } |
679 | } |
680 | |
681 | static inline void __ceph_dir_set_complete(struct ceph_inode_info *ci, |
682 | long long release_count, |
683 | long long ordered_count) |
684 | { |
685 | /* |
686 | * Makes sure operations that setup readdir cache (update page |
687 | * cache and i_size) are strongly ordered w.r.t. the following |
688 | * atomic64_set() operations. |
689 | */ |
690 | smp_mb(); |
691 | atomic64_set(v: &ci->i_complete_seq[0], i: release_count); |
692 | atomic64_set(v: &ci->i_complete_seq[1], i: ordered_count); |
693 | } |
694 | |
695 | static inline void __ceph_dir_clear_complete(struct ceph_inode_info *ci) |
696 | { |
697 | atomic64_inc(v: &ci->i_release_count); |
698 | } |
699 | |
700 | static inline void __ceph_dir_clear_ordered(struct ceph_inode_info *ci) |
701 | { |
702 | atomic64_inc(v: &ci->i_ordered_count); |
703 | } |
704 | |
705 | static inline bool __ceph_dir_is_complete(struct ceph_inode_info *ci) |
706 | { |
707 | return atomic64_read(v: &ci->i_complete_seq[0]) == |
708 | atomic64_read(v: &ci->i_release_count); |
709 | } |
710 | |
711 | static inline bool __ceph_dir_is_complete_ordered(struct ceph_inode_info *ci) |
712 | { |
713 | return atomic64_read(v: &ci->i_complete_seq[0]) == |
714 | atomic64_read(v: &ci->i_release_count) && |
715 | atomic64_read(v: &ci->i_complete_seq[1]) == |
716 | atomic64_read(v: &ci->i_ordered_count); |
717 | } |
718 | |
719 | static inline void ceph_dir_clear_complete(struct inode *inode) |
720 | { |
721 | __ceph_dir_clear_complete(ci: ceph_inode(inode)); |
722 | } |
723 | |
724 | static inline void ceph_dir_clear_ordered(struct inode *inode) |
725 | { |
726 | __ceph_dir_clear_ordered(ci: ceph_inode(inode)); |
727 | } |
728 | |
729 | static inline bool ceph_dir_is_complete_ordered(struct inode *inode) |
730 | { |
731 | bool ret = __ceph_dir_is_complete_ordered(ci: ceph_inode(inode)); |
732 | smp_rmb(); |
733 | return ret; |
734 | } |
735 | |
736 | /* find a specific frag @f */ |
737 | extern struct ceph_inode_frag *__ceph_find_frag(struct ceph_inode_info *ci, |
738 | u32 f); |
739 | |
740 | /* |
741 | * choose fragment for value @v. copy frag content to pfrag, if leaf |
742 | * exists |
743 | */ |
744 | extern u32 ceph_choose_frag(struct ceph_inode_info *ci, u32 v, |
745 | struct ceph_inode_frag *pfrag, |
746 | int *found); |
747 | |
748 | static inline struct ceph_dentry_info *ceph_dentry(const struct dentry *dentry) |
749 | { |
750 | return (struct ceph_dentry_info *)dentry->d_fsdata; |
751 | } |
752 | |
753 | /* |
754 | * caps helpers |
755 | */ |
756 | static inline bool __ceph_is_any_real_caps(struct ceph_inode_info *ci) |
757 | { |
758 | return !RB_EMPTY_ROOT(&ci->i_caps); |
759 | } |
760 | |
761 | extern int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented); |
762 | extern int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int t); |
763 | extern int __ceph_caps_issued_mask_metric(struct ceph_inode_info *ci, int mask, |
764 | int t); |
765 | extern int __ceph_caps_issued_other(struct ceph_inode_info *ci, |
766 | struct ceph_cap *cap); |
767 | |
768 | static inline int ceph_caps_issued(struct ceph_inode_info *ci) |
769 | { |
770 | int issued; |
771 | spin_lock(lock: &ci->i_ceph_lock); |
772 | issued = __ceph_caps_issued(ci, NULL); |
773 | spin_unlock(lock: &ci->i_ceph_lock); |
774 | return issued; |
775 | } |
776 | |
777 | static inline int ceph_caps_issued_mask_metric(struct ceph_inode_info *ci, |
778 | int mask, int touch) |
779 | { |
780 | int r; |
781 | spin_lock(lock: &ci->i_ceph_lock); |
782 | r = __ceph_caps_issued_mask_metric(ci, mask, t: touch); |
783 | spin_unlock(lock: &ci->i_ceph_lock); |
784 | return r; |
785 | } |
786 | |
787 | static inline int __ceph_caps_dirty(struct ceph_inode_info *ci) |
788 | { |
789 | return ci->i_dirty_caps | ci->i_flushing_caps; |
790 | } |
791 | extern struct ceph_cap_flush *ceph_alloc_cap_flush(void); |
792 | extern void ceph_free_cap_flush(struct ceph_cap_flush *cf); |
793 | extern int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask, |
794 | struct ceph_cap_flush **pcf); |
795 | |
796 | extern int __ceph_caps_revoking_other(struct ceph_inode_info *ci, |
797 | struct ceph_cap *ocap, int mask); |
798 | extern int ceph_caps_revoking(struct ceph_inode_info *ci, int mask); |
799 | extern int __ceph_caps_used(struct ceph_inode_info *ci); |
800 | |
801 | static inline bool __ceph_is_file_opened(struct ceph_inode_info *ci) |
802 | { |
803 | return ci->i_nr_by_mode[0]; |
804 | } |
805 | extern int __ceph_caps_file_wanted(struct ceph_inode_info *ci); |
806 | extern int __ceph_caps_wanted(struct ceph_inode_info *ci); |
807 | |
808 | /* what the mds thinks we want */ |
809 | extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci, bool check); |
810 | |
811 | extern void ceph_caps_init(struct ceph_mds_client *mdsc); |
812 | extern void ceph_caps_finalize(struct ceph_mds_client *mdsc); |
813 | extern void ceph_adjust_caps_max_min(struct ceph_mds_client *mdsc, |
814 | struct ceph_mount_options *fsopt); |
815 | extern int ceph_reserve_caps(struct ceph_mds_client *mdsc, |
816 | struct ceph_cap_reservation *ctx, int need); |
817 | extern void ceph_unreserve_caps(struct ceph_mds_client *mdsc, |
818 | struct ceph_cap_reservation *ctx); |
819 | extern void ceph_reservation_status(struct ceph_fs_client *client, |
820 | int *total, int *avail, int *used, |
821 | int *reserved, int *min); |
822 | extern void change_auth_cap_ses(struct ceph_inode_info *ci, |
823 | struct ceph_mds_session *session); |
824 | |
825 | |
826 | |
827 | /* |
828 | * we keep buffered readdir results attached to file->private_data |
829 | */ |
830 | #define CEPH_F_SYNC 1 |
831 | #define CEPH_F_ATEND 2 |
832 | |
833 | struct ceph_file_info { |
834 | short fmode; /* initialized on open */ |
835 | short flags; /* CEPH_F_* */ |
836 | |
837 | spinlock_t rw_contexts_lock; |
838 | struct list_head rw_contexts; |
839 | |
840 | u32 filp_gen; |
841 | }; |
842 | |
843 | struct ceph_dir_file_info { |
844 | struct ceph_file_info file_info; |
845 | |
846 | /* readdir: position within the dir */ |
847 | u32 frag; |
848 | struct ceph_mds_request *last_readdir; |
849 | |
850 | /* readdir: position within a frag */ |
851 | unsigned next_offset; /* offset of next chunk (last_name's + 1) */ |
852 | char *last_name; /* last entry in previous chunk */ |
853 | long long dir_release_count; |
854 | long long dir_ordered_count; |
855 | int readdir_cache_idx; |
856 | |
857 | /* used for -o dirstat read() on directory thing */ |
858 | char *dir_info; |
859 | int dir_info_len; |
860 | }; |
861 | |
862 | struct ceph_rw_context { |
863 | struct list_head list; |
864 | struct task_struct *thread; |
865 | int caps; |
866 | }; |
867 | |
868 | #define CEPH_DEFINE_RW_CONTEXT(_name, _caps) \ |
869 | struct ceph_rw_context _name = { \ |
870 | .thread = current, \ |
871 | .caps = _caps, \ |
872 | } |
873 | |
874 | static inline void ceph_add_rw_context(struct ceph_file_info *cf, |
875 | struct ceph_rw_context *ctx) |
876 | { |
877 | spin_lock(lock: &cf->rw_contexts_lock); |
878 | list_add(new: &ctx->list, head: &cf->rw_contexts); |
879 | spin_unlock(lock: &cf->rw_contexts_lock); |
880 | } |
881 | |
882 | static inline void ceph_del_rw_context(struct ceph_file_info *cf, |
883 | struct ceph_rw_context *ctx) |
884 | { |
885 | spin_lock(lock: &cf->rw_contexts_lock); |
886 | list_del(entry: &ctx->list); |
887 | spin_unlock(lock: &cf->rw_contexts_lock); |
888 | } |
889 | |
890 | static inline struct ceph_rw_context* |
891 | ceph_find_rw_context(struct ceph_file_info *cf) |
892 | { |
893 | struct ceph_rw_context *ctx, *found = NULL; |
894 | spin_lock(lock: &cf->rw_contexts_lock); |
895 | list_for_each_entry(ctx, &cf->rw_contexts, list) { |
896 | if (ctx->thread == current) { |
897 | found = ctx; |
898 | break; |
899 | } |
900 | } |
901 | spin_unlock(lock: &cf->rw_contexts_lock); |
902 | return found; |
903 | } |
904 | |
905 | struct ceph_readdir_cache_control { |
906 | struct page *page; |
907 | struct dentry **dentries; |
908 | int index; |
909 | }; |
910 | |
911 | /* |
912 | * A "snap realm" describes a subset of the file hierarchy sharing |
913 | * the same set of snapshots that apply to it. The realms themselves |
914 | * are organized into a hierarchy, such that children inherit (some of) |
915 | * the snapshots of their parents. |
916 | * |
917 | * All inodes within the realm that have capabilities are linked into a |
918 | * per-realm list. |
919 | */ |
920 | struct ceph_snap_realm { |
921 | u64 ino; |
922 | struct inode *inode; |
923 | atomic_t nref; |
924 | struct rb_node node; |
925 | |
926 | u64 created, seq; |
927 | u64 parent_ino; |
928 | u64 parent_since; /* snapid when our current parent became so */ |
929 | |
930 | u64 *prior_parent_snaps; /* snaps inherited from any parents we */ |
931 | u32 num_prior_parent_snaps; /* had prior to parent_since */ |
932 | u64 *snaps; /* snaps specific to this realm */ |
933 | u32 num_snaps; |
934 | |
935 | struct ceph_snap_realm *parent; |
936 | struct list_head children; /* list of child realms */ |
937 | struct list_head child_item; |
938 | |
939 | struct list_head empty_item; /* if i have ref==0 */ |
940 | |
941 | struct list_head dirty_item; /* if realm needs new context */ |
942 | |
943 | struct list_head rebuild_item; /* rebuild snap realms _downward_ in hierarchy */ |
944 | |
945 | /* the current set of snaps for this realm */ |
946 | struct ceph_snap_context *cached_context; |
947 | |
948 | struct list_head inodes_with_caps; |
949 | spinlock_t inodes_with_caps_lock; |
950 | }; |
951 | |
952 | static inline int default_congestion_kb(void) |
953 | { |
954 | int congestion_kb; |
955 | |
956 | /* |
957 | * Copied from NFS |
958 | * |
959 | * congestion size, scale with available memory. |
960 | * |
961 | * 64MB: 8192k |
962 | * 128MB: 11585k |
963 | * 256MB: 16384k |
964 | * 512MB: 23170k |
965 | * 1GB: 32768k |
966 | * 2GB: 46340k |
967 | * 4GB: 65536k |
968 | * 8GB: 92681k |
969 | * 16GB: 131072k |
970 | * |
971 | * This allows larger machines to have larger/more transfers. |
972 | * Limit the default to 256M |
973 | */ |
974 | congestion_kb = (16*int_sqrt(totalram_pages())) << (PAGE_SHIFT-10); |
975 | if (congestion_kb > 256*1024) |
976 | congestion_kb = 256*1024; |
977 | |
978 | return congestion_kb; |
979 | } |
980 | |
981 | |
982 | /* super.c */ |
983 | extern int ceph_force_reconnect(struct super_block *sb); |
984 | /* snap.c */ |
985 | struct ceph_snap_realm *ceph_lookup_snap_realm(struct ceph_mds_client *mdsc, |
986 | u64 ino); |
987 | extern void ceph_get_snap_realm(struct ceph_mds_client *mdsc, |
988 | struct ceph_snap_realm *realm); |
989 | extern void ceph_put_snap_realm(struct ceph_mds_client *mdsc, |
990 | struct ceph_snap_realm *realm); |
991 | extern int ceph_update_snap_trace(struct ceph_mds_client *m, |
992 | void *p, void *e, bool deletion, |
993 | struct ceph_snap_realm **realm_ret); |
994 | void ceph_change_snap_realm(struct inode *inode, struct ceph_snap_realm *realm); |
995 | extern void ceph_handle_snap(struct ceph_mds_client *mdsc, |
996 | struct ceph_mds_session *session, |
997 | struct ceph_msg *msg); |
998 | extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci, |
999 | struct ceph_cap_snap *capsnap); |
1000 | extern void ceph_cleanup_global_and_empty_realms(struct ceph_mds_client *mdsc); |
1001 | |
1002 | extern struct ceph_snapid_map *ceph_get_snapid_map(struct ceph_mds_client *mdsc, |
1003 | u64 snap); |
1004 | extern void ceph_put_snapid_map(struct ceph_mds_client* mdsc, |
1005 | struct ceph_snapid_map *sm); |
1006 | extern void ceph_trim_snapid_map(struct ceph_mds_client *mdsc); |
1007 | extern void ceph_cleanup_snapid_map(struct ceph_mds_client *mdsc); |
1008 | void ceph_umount_begin(struct super_block *sb); |
1009 | |
1010 | |
1011 | /* |
1012 | * a cap_snap is "pending" if it is still awaiting an in-progress |
1013 | * sync write (that may/may not still update size, mtime, etc.). |
1014 | */ |
1015 | static inline bool __ceph_have_pending_cap_snap(struct ceph_inode_info *ci) |
1016 | { |
1017 | return !list_empty(head: &ci->i_cap_snaps) && |
1018 | list_last_entry(&ci->i_cap_snaps, struct ceph_cap_snap, |
1019 | ci_item)->writing; |
1020 | } |
1021 | |
1022 | /* inode.c */ |
1023 | struct ceph_mds_reply_info_in; |
1024 | struct ceph_mds_reply_dirfrag; |
1025 | struct ceph_acl_sec_ctx; |
1026 | |
1027 | extern const struct inode_operations ceph_file_iops; |
1028 | |
1029 | extern struct inode *ceph_alloc_inode(struct super_block *sb); |
1030 | extern void ceph_evict_inode(struct inode *inode); |
1031 | extern void ceph_free_inode(struct inode *inode); |
1032 | |
1033 | struct inode *ceph_new_inode(struct inode *dir, struct dentry *dentry, |
1034 | umode_t *mode, struct ceph_acl_sec_ctx *as_ctx); |
1035 | void ceph_as_ctx_to_req(struct ceph_mds_request *req, |
1036 | struct ceph_acl_sec_ctx *as_ctx); |
1037 | |
1038 | extern struct inode *ceph_get_inode(struct super_block *sb, |
1039 | struct ceph_vino vino, |
1040 | struct inode *newino); |
1041 | extern struct inode *ceph_get_snapdir(struct inode *parent); |
1042 | extern int ceph_fill_file_size(struct inode *inode, int issued, |
1043 | u32 truncate_seq, u64 truncate_size, u64 size); |
1044 | extern void ceph_fill_file_time(struct inode *inode, int issued, |
1045 | u64 time_warp_seq, struct timespec64 *ctime, |
1046 | struct timespec64 *mtime, |
1047 | struct timespec64 *atime); |
1048 | extern int ceph_fill_inode(struct inode *inode, struct page *locked_page, |
1049 | struct ceph_mds_reply_info_in *iinfo, |
1050 | struct ceph_mds_reply_dirfrag *dirinfo, |
1051 | struct ceph_mds_session *session, int cap_fmode, |
1052 | struct ceph_cap_reservation *caps_reservation); |
1053 | extern int ceph_fill_trace(struct super_block *sb, |
1054 | struct ceph_mds_request *req); |
1055 | extern int ceph_readdir_prepopulate(struct ceph_mds_request *req, |
1056 | struct ceph_mds_session *session); |
1057 | |
1058 | extern int ceph_inode_holds_cap(struct inode *inode, int mask); |
1059 | |
1060 | extern bool ceph_inode_set_size(struct inode *inode, loff_t size); |
1061 | extern void __ceph_do_pending_vmtruncate(struct inode *inode); |
1062 | |
1063 | void ceph_queue_inode_work(struct inode *inode, int work_bit); |
1064 | |
1065 | static inline void ceph_queue_vmtruncate(struct inode *inode) |
1066 | { |
1067 | ceph_queue_inode_work(inode, CEPH_I_WORK_VMTRUNCATE); |
1068 | } |
1069 | |
1070 | static inline void ceph_queue_invalidate(struct inode *inode) |
1071 | { |
1072 | ceph_queue_inode_work(inode, CEPH_I_WORK_INVALIDATE_PAGES); |
1073 | } |
1074 | |
1075 | static inline void ceph_queue_writeback(struct inode *inode) |
1076 | { |
1077 | ceph_queue_inode_work(inode, CEPH_I_WORK_WRITEBACK); |
1078 | } |
1079 | |
1080 | static inline void ceph_queue_check_caps(struct inode *inode) |
1081 | { |
1082 | ceph_queue_inode_work(inode, CEPH_I_WORK_CHECK_CAPS); |
1083 | } |
1084 | |
1085 | static inline void ceph_queue_flush_snaps(struct inode *inode) |
1086 | { |
1087 | ceph_queue_inode_work(inode, CEPH_I_WORK_FLUSH_SNAPS); |
1088 | } |
1089 | |
1090 | extern int ceph_try_to_choose_auth_mds(struct inode *inode, int mask); |
1091 | extern int __ceph_do_getattr(struct inode *inode, struct page *locked_page, |
1092 | int mask, bool force); |
1093 | static inline int ceph_do_getattr(struct inode *inode, int mask, bool force) |
1094 | { |
1095 | return __ceph_do_getattr(inode, NULL, mask, force); |
1096 | } |
1097 | extern int ceph_permission(struct mnt_idmap *idmap, |
1098 | struct inode *inode, int mask); |
1099 | |
1100 | struct ceph_iattr { |
1101 | struct ceph_fscrypt_auth *fscrypt_auth; |
1102 | }; |
1103 | |
1104 | extern int __ceph_setattr(struct mnt_idmap *idmap, struct inode *inode, |
1105 | struct iattr *attr, struct ceph_iattr *cia); |
1106 | extern int ceph_setattr(struct mnt_idmap *idmap, |
1107 | struct dentry *dentry, struct iattr *attr); |
1108 | extern int ceph_getattr(struct mnt_idmap *idmap, |
1109 | const struct path *path, struct kstat *stat, |
1110 | u32 request_mask, unsigned int flags); |
1111 | void ceph_inode_shutdown(struct inode *inode); |
1112 | |
1113 | static inline bool ceph_inode_is_shutdown(struct inode *inode) |
1114 | { |
1115 | unsigned long flags = READ_ONCE(ceph_inode(inode)->i_ceph_flags); |
1116 | struct ceph_fs_client *fsc = ceph_inode_to_fs_client(inode); |
1117 | int state = READ_ONCE(fsc->mount_state); |
1118 | |
1119 | return (flags & CEPH_I_SHUTDOWN) || state >= CEPH_MOUNT_SHUTDOWN; |
1120 | } |
1121 | |
1122 | /* xattr.c */ |
1123 | int __ceph_setxattr(struct inode *, const char *, const void *, size_t, int); |
1124 | int ceph_do_getvxattr(struct inode *inode, const char *name, void *value, size_t size); |
1125 | ssize_t __ceph_getxattr(struct inode *, const char *, void *, size_t); |
1126 | extern ssize_t ceph_listxattr(struct dentry *, char *, size_t); |
1127 | extern struct ceph_buffer *__ceph_build_xattrs_blob(struct ceph_inode_info *ci); |
1128 | extern void __ceph_destroy_xattrs(struct ceph_inode_info *ci); |
1129 | extern const struct xattr_handler * const ceph_xattr_handlers[]; |
1130 | |
1131 | struct ceph_acl_sec_ctx { |
1132 | #ifdef CONFIG_CEPH_FS_POSIX_ACL |
1133 | void *default_acl; |
1134 | void *acl; |
1135 | #endif |
1136 | #ifdef CONFIG_CEPH_FS_SECURITY_LABEL |
1137 | void *sec_ctx; |
1138 | u32 sec_ctxlen; |
1139 | #endif |
1140 | #ifdef CONFIG_FS_ENCRYPTION |
1141 | struct ceph_fscrypt_auth *fscrypt_auth; |
1142 | #endif |
1143 | struct ceph_pagelist *pagelist; |
1144 | }; |
1145 | |
1146 | #ifdef CONFIG_SECURITY |
1147 | extern bool ceph_security_xattr_deadlock(struct inode *in); |
1148 | extern bool ceph_security_xattr_wanted(struct inode *in); |
1149 | #else |
1150 | static inline bool ceph_security_xattr_deadlock(struct inode *in) |
1151 | { |
1152 | return false; |
1153 | } |
1154 | static inline bool ceph_security_xattr_wanted(struct inode *in) |
1155 | { |
1156 | return false; |
1157 | } |
1158 | #endif |
1159 | |
1160 | #ifdef CONFIG_CEPH_FS_SECURITY_LABEL |
1161 | extern int ceph_security_init_secctx(struct dentry *dentry, umode_t mode, |
1162 | struct ceph_acl_sec_ctx *ctx); |
1163 | static inline void ceph_security_invalidate_secctx(struct inode *inode) |
1164 | { |
1165 | security_inode_invalidate_secctx(inode); |
1166 | } |
1167 | #else |
1168 | static inline int ceph_security_init_secctx(struct dentry *dentry, umode_t mode, |
1169 | struct ceph_acl_sec_ctx *ctx) |
1170 | { |
1171 | return 0; |
1172 | } |
1173 | static inline void ceph_security_invalidate_secctx(struct inode *inode) |
1174 | { |
1175 | } |
1176 | #endif |
1177 | |
1178 | void ceph_release_acl_sec_ctx(struct ceph_acl_sec_ctx *as_ctx); |
1179 | |
1180 | /* acl.c */ |
1181 | #ifdef CONFIG_CEPH_FS_POSIX_ACL |
1182 | |
1183 | struct posix_acl *ceph_get_acl(struct inode *, int, bool); |
1184 | int ceph_set_acl(struct mnt_idmap *idmap, |
1185 | struct dentry *dentry, struct posix_acl *acl, int type); |
1186 | int ceph_pre_init_acls(struct inode *dir, umode_t *mode, |
1187 | struct ceph_acl_sec_ctx *as_ctx); |
1188 | void ceph_init_inode_acls(struct inode *inode, |
1189 | struct ceph_acl_sec_ctx *as_ctx); |
1190 | |
1191 | static inline void ceph_forget_all_cached_acls(struct inode *inode) |
1192 | { |
1193 | forget_all_cached_acls(inode); |
1194 | } |
1195 | |
1196 | #else |
1197 | |
1198 | #define ceph_get_acl NULL |
1199 | #define ceph_set_acl NULL |
1200 | |
1201 | static inline int ceph_pre_init_acls(struct inode *dir, umode_t *mode, |
1202 | struct ceph_acl_sec_ctx *as_ctx) |
1203 | { |
1204 | return 0; |
1205 | } |
1206 | static inline void ceph_init_inode_acls(struct inode *inode, |
1207 | struct ceph_acl_sec_ctx *as_ctx) |
1208 | { |
1209 | } |
1210 | static inline int ceph_acl_chmod(struct dentry *dentry, struct inode *inode) |
1211 | { |
1212 | return 0; |
1213 | } |
1214 | |
1215 | static inline void ceph_forget_all_cached_acls(struct inode *inode) |
1216 | { |
1217 | } |
1218 | |
1219 | #endif |
1220 | |
1221 | /* caps.c */ |
1222 | extern const char *ceph_cap_string(int c); |
1223 | extern void ceph_handle_caps(struct ceph_mds_session *session, |
1224 | struct ceph_msg *msg); |
1225 | extern struct ceph_cap *ceph_get_cap(struct ceph_mds_client *mdsc, |
1226 | struct ceph_cap_reservation *ctx); |
1227 | extern void ceph_add_cap(struct inode *inode, |
1228 | struct ceph_mds_session *session, u64 cap_id, |
1229 | unsigned issued, unsigned wanted, |
1230 | unsigned cap, unsigned seq, u64 realmino, int flags, |
1231 | struct ceph_cap **new_cap); |
1232 | extern void __ceph_remove_cap(struct ceph_cap *cap, bool queue_release); |
1233 | extern void ceph_remove_cap(struct ceph_mds_client *mdsc, struct ceph_cap *cap, |
1234 | bool queue_release); |
1235 | extern void __ceph_remove_caps(struct ceph_inode_info *ci); |
1236 | extern void ceph_put_cap(struct ceph_mds_client *mdsc, |
1237 | struct ceph_cap *cap); |
1238 | extern int ceph_is_any_caps(struct inode *inode); |
1239 | |
1240 | extern int ceph_write_inode(struct inode *inode, struct writeback_control *wbc); |
1241 | extern int ceph_fsync(struct file *file, loff_t start, loff_t end, |
1242 | int datasync); |
1243 | extern void ceph_early_kick_flushing_caps(struct ceph_mds_client *mdsc, |
1244 | struct ceph_mds_session *session); |
1245 | extern void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc, |
1246 | struct ceph_mds_session *session); |
1247 | void ceph_kick_flushing_inode_caps(struct ceph_mds_session *session, |
1248 | struct ceph_inode_info *ci); |
1249 | extern struct ceph_cap *__get_cap_for_mds(struct ceph_inode_info *ci, |
1250 | int mds); |
1251 | extern struct ceph_cap *ceph_get_cap_for_mds(struct ceph_inode_info *ci, |
1252 | int mds); |
1253 | extern void ceph_take_cap_refs(struct ceph_inode_info *ci, int caps, |
1254 | bool snap_rwsem_locked); |
1255 | extern void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps); |
1256 | extern void ceph_put_cap_refs(struct ceph_inode_info *ci, int had); |
1257 | extern void ceph_put_cap_refs_async(struct ceph_inode_info *ci, int had); |
1258 | extern void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr, |
1259 | struct ceph_snap_context *snapc); |
1260 | extern void __ceph_remove_capsnap(struct inode *inode, |
1261 | struct ceph_cap_snap *capsnap, |
1262 | bool *wake_ci, bool *wake_mdsc); |
1263 | extern void ceph_remove_capsnap(struct inode *inode, |
1264 | struct ceph_cap_snap *capsnap, |
1265 | bool *wake_ci, bool *wake_mdsc); |
1266 | extern void ceph_flush_snaps(struct ceph_inode_info *ci, |
1267 | struct ceph_mds_session **psession); |
1268 | extern bool __ceph_should_report_size(struct ceph_inode_info *ci); |
1269 | extern void ceph_check_caps(struct ceph_inode_info *ci, int flags); |
1270 | extern unsigned long ceph_check_delayed_caps(struct ceph_mds_client *mdsc); |
1271 | extern void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc); |
1272 | extern int ceph_drop_caps_for_unlink(struct inode *inode); |
1273 | extern int ceph_encode_inode_release(void **p, struct inode *inode, |
1274 | int mds, int drop, int unless, int force); |
1275 | extern int ceph_encode_dentry_release(void **p, struct dentry *dn, |
1276 | struct inode *dir, |
1277 | int mds, int drop, int unless); |
1278 | |
1279 | extern int __ceph_get_caps(struct inode *inode, struct ceph_file_info *fi, |
1280 | int need, int want, loff_t endoff, int *got); |
1281 | extern int ceph_get_caps(struct file *filp, int need, int want, |
1282 | loff_t endoff, int *got); |
1283 | extern int ceph_try_get_caps(struct inode *inode, |
1284 | int need, int want, bool nonblock, int *got); |
1285 | |
1286 | /* for counting open files by mode */ |
1287 | extern void ceph_get_fmode(struct ceph_inode_info *ci, int mode, int count); |
1288 | extern void ceph_put_fmode(struct ceph_inode_info *ci, int mode, int count); |
1289 | extern void __ceph_touch_fmode(struct ceph_inode_info *ci, |
1290 | struct ceph_mds_client *mdsc, int fmode); |
1291 | |
1292 | /* addr.c */ |
1293 | extern const struct address_space_operations ceph_aops; |
1294 | extern const struct netfs_request_ops ceph_netfs_ops; |
1295 | extern int ceph_mmap(struct file *file, struct vm_area_struct *vma); |
1296 | extern int ceph_uninline_data(struct file *file); |
1297 | extern int ceph_pool_perm_check(struct inode *inode, int need); |
1298 | extern void ceph_pool_perm_destroy(struct ceph_mds_client* mdsc); |
1299 | int ceph_purge_inode_cap(struct inode *inode, struct ceph_cap *cap, bool *invalidate); |
1300 | |
1301 | static inline bool ceph_has_inline_data(struct ceph_inode_info *ci) |
1302 | { |
1303 | if (ci->i_inline_version == CEPH_INLINE_NONE || |
1304 | ci->i_inline_version == 1) /* initial version, no data */ |
1305 | return false; |
1306 | return true; |
1307 | } |
1308 | |
1309 | /* file.c */ |
1310 | extern const struct file_operations ceph_file_fops; |
1311 | |
1312 | extern int ceph_renew_caps(struct inode *inode, int fmode); |
1313 | extern int ceph_open(struct inode *inode, struct file *file); |
1314 | extern int ceph_atomic_open(struct inode *dir, struct dentry *dentry, |
1315 | struct file *file, unsigned flags, umode_t mode); |
1316 | extern ssize_t __ceph_sync_read(struct inode *inode, loff_t *ki_pos, |
1317 | struct iov_iter *to, int *retry_op, |
1318 | u64 *last_objver); |
1319 | extern int ceph_release(struct inode *inode, struct file *filp); |
1320 | extern void ceph_fill_inline_data(struct inode *inode, struct page *locked_page, |
1321 | char *data, size_t len); |
1322 | |
1323 | /* dir.c */ |
1324 | extern const struct file_operations ceph_dir_fops; |
1325 | extern const struct file_operations ceph_snapdir_fops; |
1326 | extern const struct inode_operations ceph_dir_iops; |
1327 | extern const struct inode_operations ceph_snapdir_iops; |
1328 | extern const struct dentry_operations ceph_dentry_ops; |
1329 | |
1330 | extern loff_t ceph_make_fpos(unsigned high, unsigned off, bool hash_order); |
1331 | extern int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry); |
1332 | extern struct dentry *ceph_handle_snapdir(struct ceph_mds_request *req, |
1333 | struct dentry *dentry); |
1334 | extern struct dentry *ceph_finish_lookup(struct ceph_mds_request *req, |
1335 | struct dentry *dentry, int err); |
1336 | |
1337 | extern void __ceph_dentry_lease_touch(struct ceph_dentry_info *di); |
1338 | extern void __ceph_dentry_dir_lease_touch(struct ceph_dentry_info *di); |
1339 | extern void ceph_invalidate_dentry_lease(struct dentry *dentry); |
1340 | extern int ceph_trim_dentries(struct ceph_mds_client *mdsc); |
1341 | extern unsigned ceph_dentry_hash(struct inode *dir, struct dentry *dn); |
1342 | extern void ceph_readdir_cache_release(struct ceph_readdir_cache_control *ctl); |
1343 | |
1344 | /* ioctl.c */ |
1345 | extern long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg); |
1346 | |
1347 | /* export.c */ |
1348 | extern const struct export_operations ceph_export_ops; |
1349 | struct inode *ceph_lookup_inode(struct super_block *sb, u64 ino); |
1350 | |
1351 | /* locks.c */ |
1352 | extern __init void ceph_flock_init(void); |
1353 | extern int ceph_lock(struct file *file, int cmd, struct file_lock *fl); |
1354 | extern int ceph_flock(struct file *file, int cmd, struct file_lock *fl); |
1355 | extern void ceph_count_locks(struct inode *inode, int *p_num, int *f_num); |
1356 | extern int ceph_encode_locks_to_buffer(struct inode *inode, |
1357 | struct ceph_filelock *flocks, |
1358 | int num_fcntl_locks, |
1359 | int num_flock_locks); |
1360 | extern int ceph_locks_to_pagelist(struct ceph_filelock *flocks, |
1361 | struct ceph_pagelist *pagelist, |
1362 | int num_fcntl_locks, int num_flock_locks); |
1363 | |
1364 | /* debugfs.c */ |
1365 | extern void ceph_fs_debugfs_init(struct ceph_fs_client *client); |
1366 | extern void ceph_fs_debugfs_cleanup(struct ceph_fs_client *client); |
1367 | |
1368 | /* quota.c */ |
1369 | |
1370 | enum quota_get_realm { |
1371 | QUOTA_GET_MAX_FILES, |
1372 | QUOTA_GET_MAX_BYTES, |
1373 | QUOTA_GET_ANY |
1374 | }; |
1375 | |
1376 | static inline bool __ceph_has_quota(struct ceph_inode_info *ci, |
1377 | enum quota_get_realm which) |
1378 | { |
1379 | bool has_quota = false; |
1380 | |
1381 | switch (which) { |
1382 | case QUOTA_GET_MAX_BYTES: |
1383 | has_quota = !!ci->i_max_bytes; |
1384 | break; |
1385 | case QUOTA_GET_MAX_FILES: |
1386 | has_quota = !!ci->i_max_files; |
1387 | break; |
1388 | default: |
1389 | has_quota = !!(ci->i_max_files || ci->i_max_bytes); |
1390 | } |
1391 | return has_quota; |
1392 | } |
1393 | |
1394 | extern void ceph_adjust_quota_realms_count(struct inode *inode, bool inc); |
1395 | |
1396 | static inline void __ceph_update_quota(struct ceph_inode_info *ci, |
1397 | u64 max_bytes, u64 max_files) |
1398 | { |
1399 | bool had_quota, has_quota; |
1400 | had_quota = __ceph_has_quota(ci, which: QUOTA_GET_ANY); |
1401 | ci->i_max_bytes = max_bytes; |
1402 | ci->i_max_files = max_files; |
1403 | has_quota = __ceph_has_quota(ci, which: QUOTA_GET_ANY); |
1404 | |
1405 | if (had_quota != has_quota) |
1406 | ceph_adjust_quota_realms_count(inode: &ci->netfs.inode, inc: has_quota); |
1407 | } |
1408 | |
1409 | static inline int __ceph_sparse_read_ext_count(struct inode *inode, u64 len) |
1410 | { |
1411 | int cnt = 0; |
1412 | |
1413 | if (IS_ENCRYPTED(inode)) { |
1414 | cnt = len >> CEPH_FSCRYPT_BLOCK_SHIFT; |
1415 | if (cnt > CEPH_SPARSE_EXT_ARRAY_INITIAL) |
1416 | cnt = 0; |
1417 | } |
1418 | |
1419 | return cnt; |
1420 | } |
1421 | |
1422 | extern void ceph_handle_quota(struct ceph_mds_client *mdsc, |
1423 | struct ceph_mds_session *session, |
1424 | struct ceph_msg *msg); |
1425 | extern bool ceph_quota_is_max_files_exceeded(struct inode *inode); |
1426 | extern bool ceph_quota_is_same_realm(struct inode *old, struct inode *new); |
1427 | extern bool ceph_quota_is_max_bytes_exceeded(struct inode *inode, |
1428 | loff_t newlen); |
1429 | extern bool ceph_quota_is_max_bytes_approaching(struct inode *inode, |
1430 | loff_t newlen); |
1431 | extern bool ceph_quota_update_statfs(struct ceph_fs_client *fsc, |
1432 | struct kstatfs *buf); |
1433 | extern void ceph_cleanup_quotarealms_inodes(struct ceph_mds_client *mdsc); |
1434 | |
1435 | bool ceph_inc_mds_stopping_blocker(struct ceph_mds_client *mdsc, |
1436 | struct ceph_mds_session *session); |
1437 | void ceph_dec_mds_stopping_blocker(struct ceph_mds_client *mdsc); |
1438 | bool ceph_inc_osd_stopping_blocker(struct ceph_mds_client *mdsc); |
1439 | void ceph_dec_osd_stopping_blocker(struct ceph_mds_client *mdsc); |
1440 | #endif /* _FS_CEPH_SUPER_H */ |
1441 | |