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 <linux/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 #define CEPH_MOUNT_OPT_NEARFULL_SYNC (1<<18) /* sync writes when nearfull */
49
50 #define CEPH_MOUNT_OPT_DEFAULT \
51 (CEPH_MOUNT_OPT_DCACHE | \
52 CEPH_MOUNT_OPT_NOCOPYFROM | \
53 CEPH_MOUNT_OPT_ASYNC_DIROPS)
54
55 #define ceph_set_mount_opt(fsc, opt) \
56 (fsc)->mount_options->flags |= CEPH_MOUNT_OPT_##opt
57 #define ceph_clear_mount_opt(fsc, opt) \
58 (fsc)->mount_options->flags &= ~CEPH_MOUNT_OPT_##opt
59 #define ceph_test_mount_opt(fsc, opt) \
60 (!!((fsc)->mount_options->flags & CEPH_MOUNT_OPT_##opt))
61
62 /* max size of osd read request, limited by libceph */
63 #define CEPH_MAX_READ_SIZE CEPH_MSG_MAX_DATA_LEN
64 /* osd has a configurable limitation of max write size.
65 * CEPH_MSG_MAX_DATA_LEN should be small enough. */
66 #define CEPH_MAX_WRITE_SIZE CEPH_MSG_MAX_DATA_LEN
67 #define CEPH_RASIZE_DEFAULT (8192*1024) /* max readahead */
68 #define CEPH_MAX_READDIR_DEFAULT 1024
69 #define CEPH_MAX_READDIR_BYTES_DEFAULT (512*1024)
70 #define CEPH_SNAPDIRNAME_DEFAULT ".snap"
71
72 /*
73 * Delay telling the MDS we no longer want caps, in case we reopen
74 * the file. Delay a minimum amount of time, even if we send a cap
75 * message for some other reason. Otherwise, take the oppotunity to
76 * update the mds to avoid sending another message later.
77 */
78 #define CEPH_CAPS_WANTED_DELAY_MIN_DEFAULT 5 /* cap release delay */
79 #define CEPH_CAPS_WANTED_DELAY_MAX_DEFAULT 60 /* cap release delay */
80
81 struct ceph_mount_options {
82 unsigned int flags;
83
84 unsigned int wsize; /* max write size */
85 unsigned int rsize; /* max read size */
86 unsigned int rasize; /* max readahead */
87 unsigned int congestion_kb; /* max writeback in flight */
88 unsigned int caps_wanted_delay_min, caps_wanted_delay_max;
89 int caps_max;
90 unsigned int max_readdir; /* max readdir result (entries) */
91 unsigned int max_readdir_bytes; /* max readdir result (bytes) */
92
93 bool new_dev_syntax;
94
95 /*
96 * everything above this point can be memcmp'd; everything below
97 * is handled in compare_mount_options()
98 */
99
100 char *snapdir_name; /* default ".snap" */
101 char *mds_namespace; /* default NULL */
102 char *server_path; /* default NULL (means "/") */
103 char *fscache_uniq; /* default NULL */
104 char *mon_addr;
105 struct fscrypt_dummy_policy dummy_enc_policy;
106 };
107
108 #define CEPH_NAMESPACE_WILDCARD "*"
109
ceph_namespace_match(const char * pattern,const char * target)110 static inline bool ceph_namespace_match(const char *pattern,
111 const char *target)
112 {
113 if (!pattern || !pattern[0] ||
114 !strcmp(pattern, CEPH_NAMESPACE_WILDCARD))
115 return true;
116
117 return !strcmp(pattern, target);
118 }
119
120 /*
121 * Check if the mds namespace in ceph_mount_options matches
122 * the passed in namespace string. First time match (when
123 * ->mds_namespace is NULL) is treated specially, since
124 * ->mds_namespace needs to be initialized by the caller.
125 */
namespace_equals(struct ceph_mount_options * fsopt,const char * namespace,size_t len)126 static inline bool namespace_equals(struct ceph_mount_options *fsopt,
127 const char *namespace, size_t len)
128 {
129 return !(fsopt->mds_namespace &&
130 (strlen(fsopt->mds_namespace) != len ||
131 strncmp(fsopt->mds_namespace, namespace, len)));
132 }
133
134 /* mount state */
135 enum {
136 CEPH_MOUNT_MOUNTING,
137 CEPH_MOUNT_MOUNTED,
138 CEPH_MOUNT_UNMOUNTING,
139 CEPH_MOUNT_UNMOUNTED,
140 CEPH_MOUNT_SHUTDOWN,
141 CEPH_MOUNT_RECOVER,
142 CEPH_MOUNT_FENCE_IO,
143 };
144
145 #define CEPH_ASYNC_CREATE_CONFLICT_BITS 8
146
147 struct ceph_fs_client {
148 struct super_block *sb;
149
150 struct list_head metric_wakeup;
151
152 struct ceph_mount_options *mount_options;
153 struct ceph_client *client;
154
155 int mount_state;
156
157 bool blocklisted;
158
159 bool have_copy_from2;
160
161 u32 filp_gen;
162 loff_t max_file_size;
163
164 struct ceph_mds_client *mdsc;
165
166 atomic_long_t writeback_count;
167 bool write_congested;
168
169 struct workqueue_struct *inode_wq;
170 struct workqueue_struct *cap_wq;
171
172 DECLARE_HASHTABLE(async_unlink_conflict, CEPH_ASYNC_CREATE_CONFLICT_BITS);
173 spinlock_t async_unlink_conflict_lock;
174
175 #ifdef CONFIG_DEBUG_FS
176 struct dentry *debugfs_dentry_lru, *debugfs_caps;
177 struct dentry *debugfs_congestion_kb;
178 struct dentry *debugfs_bdi;
179 struct dentry *debugfs_mdsc, *debugfs_mdsmap;
180 struct dentry *debugfs_status;
181 struct dentry *debugfs_mds_sessions;
182 struct dentry *debugfs_metrics_dir;
183 struct dentry *debugfs_reset_dir;
184 struct dentry *debugfs_subvolume_metrics;
185 #endif
186
187 #ifdef CONFIG_CEPH_FSCACHE
188 struct fscache_volume *fscache;
189 #endif
190 #ifdef CONFIG_FS_ENCRYPTION
191 struct fscrypt_dummy_policy fsc_dummy_enc_policy;
192 #endif
193 };
194
195 /*
196 * File i/o capability. This tracks shared state with the metadata
197 * server that allows us to cache or writeback attributes or to read
198 * and write data. For any given inode, we should have one or more
199 * capabilities, one issued by each metadata server, and our
200 * cumulative access is the OR of all issued capabilities.
201 *
202 * Each cap is referenced by the inode's i_caps rbtree and by per-mds
203 * session capability lists.
204 */
205 struct ceph_cap {
206 struct ceph_inode_info *ci;
207
208 /**
209 * Per-ci cap tree. Protected with
210 * `ceph_inode_info.i_ceph_lock`.
211 *
212 * Clearing this field with RB_CLEAR_NODE() requires holding
213 * both `ceph_inode_info.i_ceph_lock` and
214 * `ceph_mds_session->s_cap_lock`. Calling RB_EMPTY_NODE()
215 * (via ceph_cap_is_removed()) requires holding at least one
216 * of these.
217 */
218 struct rb_node ci_node;
219
220 struct ceph_mds_session *session;
221 struct list_head session_caps; /* per-session caplist */
222 u64 cap_id; /* unique cap id (mds provided) */
223 union {
224 /* in-use caps */
225 struct {
226 int issued; /* latest, from the mds */
227 int implemented; /* implemented superset of
228 issued (for revocation) */
229 int mds; /* mds index for this cap */
230 int mds_wanted; /* caps wanted from this mds */
231 };
232 /* caps to release */
233 struct {
234 u64 cap_ino;
235 int queue_release;
236 };
237 };
238 u32 seq, issue_seq, mseq;
239 u32 cap_gen; /* active/stale cycle */
240 unsigned long last_used;
241 struct list_head caps_item;
242 };
243
244 #define CHECK_CAPS_AUTHONLY 1 /* only check auth cap */
245 #define CHECK_CAPS_FLUSH 2 /* flush any dirty caps */
246 #define CHECK_CAPS_NOINVAL 4 /* don't invalidate pagecache */
247 #define CHECK_CAPS_FLUSH_FORCE 8 /* force flush any caps */
248
249 struct ceph_cap_flush {
250 u64 tid;
251 int caps;
252 bool wake; /* wake up flush waiters when finish ? */
253 bool is_capsnap; /* true means capsnap */
254 struct list_head g_list; // global
255 struct list_head i_list; // per inode
256 struct ceph_inode_info *ci;
257 };
258
259 /*
260 * Snapped cap state that is pending flush to mds. When a snapshot occurs,
261 * we first complete any in-process sync writes and writeback any dirty
262 * data before flushing the snapped state (tracked here) back to the MDS.
263 */
264 struct ceph_cap_snap {
265 refcount_t nref;
266 struct list_head ci_item;
267
268 struct ceph_cap_flush cap_flush;
269
270 u64 follows;
271 int issued, dirty;
272 struct ceph_snap_context *context;
273
274 umode_t mode;
275 kuid_t uid;
276 kgid_t gid;
277
278 struct ceph_buffer *xattr_blob;
279 u64 xattr_version;
280
281 u64 size;
282 u64 change_attr;
283 struct timespec64 mtime, atime, ctime, btime;
284 u64 time_warp_seq;
285 u64 truncate_size;
286 u32 truncate_seq;
287 int writing; /* a sync write is still in progress */
288 int dirty_pages; /* dirty pages awaiting writeback */
289 bool inline_data;
290 bool need_flush;
291 };
292
ceph_put_cap_snap(struct ceph_cap_snap * capsnap)293 static inline void ceph_put_cap_snap(struct ceph_cap_snap *capsnap)
294 {
295 if (refcount_dec_and_test(&capsnap->nref)) {
296 if (capsnap->xattr_blob)
297 ceph_buffer_put(capsnap->xattr_blob);
298 kmem_cache_free(ceph_cap_snap_cachep, capsnap);
299 }
300 }
301
302 /*
303 * The frag tree describes how a directory is fragmented, potentially across
304 * multiple metadata servers. It is also used to indicate points where
305 * metadata authority is delegated, and whether/where metadata is replicated.
306 *
307 * A _leaf_ frag will be present in the i_fragtree IFF there is
308 * delegation info. That is, if mds >= 0 || ndist > 0.
309 */
310 #define CEPH_MAX_DIRFRAG_REP 4
311
312 struct ceph_inode_frag {
313 struct rb_node node;
314
315 /* fragtree state */
316 u32 frag;
317 int split_by; /* i.e. 2^(split_by) children */
318
319 /* delegation and replication info */
320 int mds; /* -1 if same authority as parent */
321 int ndist; /* >0 if replicated */
322 int dist[CEPH_MAX_DIRFRAG_REP];
323 };
324
325 /*
326 * We cache inode xattrs as an encoded blob until they are first used,
327 * at which point we parse them into an rbtree.
328 */
329 struct ceph_inode_xattr {
330 struct rb_node node;
331
332 const char *name;
333 int name_len;
334 const char *val;
335 int val_len;
336 int dirty;
337
338 int should_free_name;
339 int should_free_val;
340 };
341
342 /*
343 * Ceph dentry state
344 */
345 struct ceph_dentry_info {
346 struct dentry *dentry;
347 struct ceph_mds_session *lease_session;
348 struct list_head lease_list;
349 struct hlist_node hnode;
350 unsigned long flags;
351 int lease_shared_gen;
352 u32 lease_gen;
353 u32 lease_seq;
354 unsigned long lease_renew_after, lease_renew_from;
355 unsigned long time;
356 u64 offset;
357 };
358
359 #define CEPH_DENTRY_REFERENCED (1 << 0)
360 #define CEPH_DENTRY_LEASE_LIST (1 << 1)
361 #define CEPH_DENTRY_SHRINK_LIST (1 << 2)
362 #define CEPH_DENTRY_PRIMARY_LINK (1 << 3)
363 #define CEPH_DENTRY_ASYNC_UNLINK_BIT (4)
364 #define CEPH_DENTRY_ASYNC_UNLINK (1 << CEPH_DENTRY_ASYNC_UNLINK_BIT)
365 #define CEPH_DENTRY_ASYNC_CREATE_BIT (5)
366 #define CEPH_DENTRY_ASYNC_CREATE (1 << CEPH_DENTRY_ASYNC_CREATE_BIT)
367
368 struct ceph_inode_xattrs_info {
369 /*
370 * (still encoded) xattr blob. we avoid the overhead of parsing
371 * this until someone actually calls getxattr, etc.
372 *
373 * blob->vec.iov_len == 4 implies there are no xattrs; blob ==
374 * NULL means we don't know.
375 */
376 struct ceph_buffer *blob, *prealloc_blob;
377
378 struct rb_root index;
379 bool dirty;
380 int count;
381 int names_size;
382 int vals_size;
383 u64 version, index_version;
384 };
385
386 /*
387 * Ceph inode.
388 */
389 struct ceph_inode_info {
390 struct netfs_inode netfs; /* Netfslib context and vfs inode */
391 struct ceph_vino i_vino; /* ceph ino + snap */
392
393 spinlock_t i_ceph_lock;
394
395 u64 i_version;
396 u64 i_inline_version;
397 u32 i_time_warp_seq;
398
399 unsigned long i_ceph_flags;
400 atomic64_t i_release_count;
401 atomic64_t i_ordered_count;
402 atomic64_t i_complete_seq[2];
403
404 struct ceph_dir_layout i_dir_layout;
405 struct ceph_file_layout i_layout;
406 struct ceph_file_layout i_cached_layout; // for async creates
407 char *i_symlink;
408
409 /* for dirs */
410 struct timespec64 i_rctime;
411 u64 i_rbytes, i_rfiles, i_rsubdirs, i_rsnaps;
412 u64 i_files, i_subdirs;
413
414 /* quotas */
415 u64 i_max_bytes, i_max_files;
416
417 /*
418 * Subvolume ID this inode belongs to. CEPH_SUBVOLUME_ID_NONE (0)
419 * means unknown/unset, matching the FUSE client convention.
420 * Once set to a valid (non-zero) value, it should not change
421 * during the inode's lifetime.
422 */
423 #define CEPH_SUBVOLUME_ID_NONE 0
424 u64 i_subvolume_id;
425
426 s32 i_dir_pin;
427
428 struct rb_root i_fragtree;
429 int i_fragtree_nsplits;
430 struct mutex i_fragtree_mutex;
431
432 struct ceph_inode_xattrs_info i_xattrs;
433
434 /* capabilities. protected _both_ by i_ceph_lock and cap->session's
435 * s_mutex. */
436 struct rb_root i_caps; /* cap list */
437 struct ceph_cap *i_auth_cap; /* authoritative cap, if any */
438 unsigned i_dirty_caps, i_flushing_caps; /* mask of dirtied fields */
439
440 /*
441 * Link to the auth cap's session's s_cap_dirty list. s_cap_dirty
442 * is protected by the mdsc->cap_dirty_lock, but each individual item
443 * is also protected by the inode's i_ceph_lock. Walking s_cap_dirty
444 * requires the mdsc->cap_dirty_lock. List presence for an item can
445 * be tested under the i_ceph_lock. Changing anything requires both.
446 */
447 struct list_head i_dirty_item;
448
449 /*
450 * Link to session's s_cap_flushing list. Protected in a similar
451 * fashion to i_dirty_item, but also by the s_mutex for changes. The
452 * s_cap_flushing list can be walked while holding either the s_mutex
453 * or msdc->cap_dirty_lock. List presence can also be checked while
454 * holding the i_ceph_lock for this inode.
455 */
456 struct list_head i_flushing_item;
457
458 /* we need to track cap writeback on a per-cap-bit basis, to allow
459 * overlapping, pipelined cap flushes to the mds. we can probably
460 * reduce the tid to 8 bits if we're concerned about inode size. */
461 struct ceph_cap_flush *i_prealloc_cap_flush;
462 struct list_head i_cap_flush_list;
463 wait_queue_head_t i_cap_wq; /* threads waiting on a capability */
464 unsigned long i_hold_caps_max; /* jiffies */
465 struct list_head i_cap_delay_list; /* for delayed cap release to mds */
466 struct ceph_cap_reservation i_cap_migration_resv;
467 struct list_head i_cap_snaps; /* snapped state pending flush to mds */
468 struct ceph_snap_context *i_head_snapc; /* set if wr_buffer_head > 0 or
469 dirty|flushing caps */
470 unsigned i_snap_caps; /* cap bits for snapped files */
471 /*
472 * Written under i_ceph_lock, read via READ_ONCE()
473 * from diagnostic paths.
474 */
475 u64 i_last_cap_flush_ack;
476
477 unsigned long i_last_rd;
478 unsigned long i_last_wr;
479 int i_nr_by_mode[CEPH_FILE_MODE_BITS]; /* open file counts */
480
481 struct mutex i_truncate_mutex;
482 u32 i_truncate_seq; /* last truncate to smaller size */
483 u64 i_truncate_size; /* and the size we last truncated down to */
484 bool i_truncate_pending; /* still need to call vmtruncate */
485 bool i_truncate_shrank; /* Truncate shrank the file */
486 /*
487 * For none fscrypt case it equals to i_truncate_size or it will
488 * equals to fscrypt_file_size
489 */
490 u64 i_truncate_pagecache_size;
491
492 u64 i_max_size; /* max file size authorized by mds */
493 u64 i_reported_size; /* (max_)size reported to or requested of mds */
494 u64 i_wanted_max_size; /* offset we'd like to write too */
495 u64 i_requested_max_size; /* max_size we've requested */
496
497 /* held references to caps */
498 int i_pin_ref;
499 int i_rd_ref, i_rdcache_ref, i_wr_ref, i_wb_ref, i_fx_ref;
500 int i_wrbuffer_ref, i_wrbuffer_ref_head;
501 atomic_t i_filelock_ref;
502 atomic_t i_shared_gen; /* increment each time we get FILE_SHARED */
503 u32 i_rdcache_gen; /* incremented each time we get FILE_CACHE. */
504 u32 i_rdcache_revoking; /* RDCACHE gen to async invalidate, if any */
505
506 struct list_head i_unsafe_dirops; /* uncommitted mds dir ops */
507 struct list_head i_unsafe_iops; /* uncommitted mds inode ops */
508 spinlock_t i_unsafe_lock;
509
510 union {
511 struct ceph_snap_realm *i_snap_realm; /* snap realm (if caps) */
512 struct ceph_snapid_map *i_snapid_map; /* snapid -> dev_t */
513 };
514 struct list_head i_snap_realm_item;
515 struct list_head i_snap_flush_item;
516 struct timespec64 i_btime;
517 struct timespec64 i_snap_btime;
518
519 struct work_struct i_work;
520 unsigned long i_work_mask;
521
522 #ifdef CONFIG_FS_ENCRYPTION
523 struct fscrypt_inode_info *i_crypt_info;
524 u32 fscrypt_auth_len;
525 u32 fscrypt_file_len;
526 u8 *fscrypt_auth;
527 u8 *fscrypt_file;
528 #endif
529 };
530
531 struct ceph_netfs_request_data {
532 int caps;
533
534 /*
535 * Maximum size of a file readahead request.
536 * The fadvise could update the bdi's default ra_pages.
537 */
538 unsigned int file_ra_pages;
539
540 /* Set it if fadvise disables file readahead entirely */
541 bool file_ra_disabled;
542 };
543
544 static inline struct ceph_inode_info *
ceph_inode(const struct inode * inode)545 ceph_inode(const struct inode *inode)
546 {
547 return container_of(inode, struct ceph_inode_info, netfs.inode);
548 }
549
550 static inline struct ceph_fs_client *
ceph_inode_to_fs_client(const struct inode * inode)551 ceph_inode_to_fs_client(const struct inode *inode)
552 {
553 return (struct ceph_fs_client *)inode->i_sb->s_fs_info;
554 }
555
556 static inline struct ceph_fs_client *
ceph_sb_to_fs_client(const struct super_block * sb)557 ceph_sb_to_fs_client(const struct super_block *sb)
558 {
559 return (struct ceph_fs_client *)sb->s_fs_info;
560 }
561
562 static inline struct ceph_mds_client *
ceph_sb_to_mdsc(const struct super_block * sb)563 ceph_sb_to_mdsc(const struct super_block *sb)
564 {
565 return (struct ceph_mds_client *)ceph_sb_to_fs_client(sb)->mdsc;
566 }
567
568 static inline struct ceph_client *
ceph_inode_to_client(const struct inode * inode)569 ceph_inode_to_client(const struct inode *inode)
570 {
571 return (struct ceph_client *)ceph_inode_to_fs_client(inode)->client;
572 }
573
574 static inline struct ceph_vino
ceph_vino(const struct inode * inode)575 ceph_vino(const struct inode *inode)
576 {
577 return ceph_inode(inode)->i_vino;
578 }
579
ceph_ino_to_ino32(u64 vino)580 static inline u32 ceph_ino_to_ino32(u64 vino)
581 {
582 u32 ino = vino & 0xffffffff;
583 ino ^= vino >> 32;
584 if (!ino)
585 ino = 2;
586 return ino;
587 }
588
589 /*
590 * Inode numbers in cephfs are 64 bits, but inode->i_ino is 32-bits on
591 * some arches. We generally do not use this value inside the ceph driver, but
592 * we do want to set it to something, so that generic vfs code has an
593 * appropriate value for tracepoints and the like.
594 */
ceph_vino_to_ino_t(struct ceph_vino vino)595 static inline ino_t ceph_vino_to_ino_t(struct ceph_vino vino)
596 {
597 if (sizeof(ino_t) == sizeof(u32))
598 return ceph_ino_to_ino32(vino.ino);
599 return (ino_t)vino.ino;
600 }
601
602 /* for printf-style formatting */
603 #define ceph_vinop(i) ceph_inode(i)->i_vino.ino, ceph_inode(i)->i_vino.snap
604
ceph_ino(struct inode * inode)605 static inline u64 ceph_ino(struct inode *inode)
606 {
607 return ceph_inode(inode)->i_vino.ino;
608 }
609
ceph_snap(struct inode * inode)610 static inline u64 ceph_snap(struct inode *inode)
611 {
612 return ceph_inode(inode)->i_vino.snap;
613 }
614
615 /**
616 * ceph_present_ino - format an inode number for presentation to userland
617 * @sb: superblock where the inode lives
618 * @ino: inode number to (possibly) convert
619 *
620 * If the user mounted with the ino32 option, then the 64-bit value needs
621 * to be converted to something that can fit inside 32 bits. Note that
622 * internal kernel code never uses this value, so this is entirely for
623 * userland consumption.
624 */
ceph_present_ino(struct super_block * sb,u64 ino)625 static inline u64 ceph_present_ino(struct super_block *sb, u64 ino)
626 {
627 if (unlikely(ceph_test_mount_opt(ceph_sb_to_fs_client(sb), INO32)))
628 return ceph_ino_to_ino32(ino);
629 return ino;
630 }
631
ceph_present_inode(struct inode * inode)632 static inline u64 ceph_present_inode(struct inode *inode)
633 {
634 return ceph_present_ino(inode->i_sb, ceph_ino(inode));
635 }
636
ceph_ino_compare(struct inode * inode,void * data)637 static inline int ceph_ino_compare(struct inode *inode, void *data)
638 {
639 struct ceph_vino *pvino = (struct ceph_vino *)data;
640 struct ceph_inode_info *ci = ceph_inode(inode);
641 return ci->i_vino.ino == pvino->ino &&
642 ci->i_vino.snap == pvino->snap;
643 }
644
645 /*
646 * The MDS reserves a set of inodes for its own usage. These should never
647 * be accessible by clients, and so the MDS has no reason to ever hand these
648 * out. The range is CEPH_MDS_INO_MDSDIR_OFFSET..CEPH_INO_SYSTEM_BASE.
649 *
650 * These come from src/mds/mdstypes.h in the ceph sources.
651 */
652 #define CEPH_MAX_MDS 0x100
653 #define CEPH_NUM_STRAY 10
654 #define CEPH_MDS_INO_MDSDIR_OFFSET (1 * CEPH_MAX_MDS)
655 #define CEPH_MDS_INO_LOG_OFFSET (2 * CEPH_MAX_MDS)
656 #define CEPH_INO_SYSTEM_BASE ((6*CEPH_MAX_MDS) + (CEPH_MAX_MDS * CEPH_NUM_STRAY))
657
658 /*
659 * Upper bound on the number of delegated inodes a single MDS session may
660 * hold. The MDS normally hands out a small preallocation window (the
661 * userspace mds_client_prealloc_inos option defaults to 1000) and refills
662 * it as the client consumes entries. This leaves generous headroom while
663 * bounding the CPU and memory a malformed delegation interval can consume.
664 */
665 #define CEPH_MAX_DELEG_INOS 8192
666
ceph_vino_is_reserved(const struct ceph_vino vino)667 static inline bool ceph_vino_is_reserved(const struct ceph_vino vino)
668 {
669 if (vino.ino >= CEPH_INO_SYSTEM_BASE ||
670 vino.ino < CEPH_MDS_INO_MDSDIR_OFFSET)
671 return false;
672
673 /* Don't warn on mdsdirs */
674 WARN_RATELIMIT(vino.ino >= CEPH_MDS_INO_LOG_OFFSET,
675 "Attempt to access reserved inode number 0x%llx",
676 vino.ino);
677 return true;
678 }
679
ceph_find_inode(struct super_block * sb,struct ceph_vino vino)680 static inline struct inode *ceph_find_inode(struct super_block *sb,
681 struct ceph_vino vino)
682 {
683 if (ceph_vino_is_reserved(vino))
684 return NULL;
685
686 /*
687 * NB: The hashval will be run through the fs/inode.c hash function
688 * anyway, so there is no need to squash the inode number down to
689 * 32-bits first. Just use low-order bits on arches with 32-bit long.
690 */
691 return ilookup5(sb, (unsigned long)vino.ino, ceph_ino_compare, &vino);
692 }
693
694
695 /*
696 * Ceph inode.
697 */
698 #define CEPH_I_DIR_ORDERED_BIT (0) /* dentries in dir are ordered */
699 /* bit 1 historically unused */
700 #define CEPH_I_FLUSH_BIT (2) /* do not delay flush of dirty metadata */
701 #define CEPH_I_POOL_PERM_BIT (3) /* pool rd/wr bits are valid */
702 #define CEPH_I_POOL_RD_BIT (4) /* can read from pool */
703 #define CEPH_I_POOL_WR_BIT (5) /* can write to pool */
704 #define CEPH_I_SEC_INITED_BIT (6) /* security initialized */
705 #define CEPH_I_KICK_FLUSH_BIT (7) /* kick flushing caps */
706 #define CEPH_I_FLUSH_SNAPS_BIT (8) /* need flush snaps */
707 #define CEPH_I_ERROR_WRITE_BIT (9) /* have seen write errors */
708 #define CEPH_I_ERROR_FILELOCK_BIT (10) /* have seen file lock errors */
709 #define CEPH_I_ODIRECT_BIT (11) /* inode in direct I/O mode */
710 #define CEPH_I_ASYNC_CREATE_BIT (12) /* async create in flight for this */
711 #define CEPH_I_SHUTDOWN_BIT (13) /* inode is no longer usable */
712 #define CEPH_I_ASYNC_CHECK_CAPS_BIT (14) /* check caps after async creating finishes */
713 #define CEPH_I_FLUSH_FORCE_BIT (15) /* a revoke's response was deferred;
714 * force a cap message to the MDS once
715 * the deferred work completes
716 */
717
718 #define CEPH_I_DIR_ORDERED (1 << CEPH_I_DIR_ORDERED_BIT)
719 #define CEPH_I_FLUSH (1 << CEPH_I_FLUSH_BIT)
720 #define CEPH_I_POOL_PERM (1 << CEPH_I_POOL_PERM_BIT)
721 #define CEPH_I_POOL_RD (1 << CEPH_I_POOL_RD_BIT)
722 #define CEPH_I_POOL_WR (1 << CEPH_I_POOL_WR_BIT)
723 #define CEPH_I_SEC_INITED (1 << CEPH_I_SEC_INITED_BIT)
724 #define CEPH_I_KICK_FLUSH (1 << CEPH_I_KICK_FLUSH_BIT)
725 #define CEPH_I_FLUSH_SNAPS (1 << CEPH_I_FLUSH_SNAPS_BIT)
726 #define CEPH_I_ODIRECT (1 << CEPH_I_ODIRECT_BIT)
727 #define CEPH_I_ASYNC_CREATE (1 << CEPH_I_ASYNC_CREATE_BIT)
728 #define CEPH_I_SHUTDOWN (1 << CEPH_I_SHUTDOWN_BIT)
729 #define CEPH_I_FLUSH_FORCE (1 << CEPH_I_FLUSH_FORCE_BIT)
730
731 /*
732 * Masks of ceph inode work.
733 */
734 #define CEPH_I_WORK_WRITEBACK 0
735 #define CEPH_I_WORK_INVALIDATE_PAGES 1
736 #define CEPH_I_WORK_VMTRUNCATE 2
737 #define CEPH_I_WORK_CHECK_CAPS 3
738 #define CEPH_I_WORK_FLUSH_SNAPS 4
739
740 /*
741 * We set the ERROR_WRITE bit when we start seeing write errors on an inode
742 * and then clear it when they start succeeding. The write submission code
743 * just takes this as a hint, so we're not too worried if a few slip through
744 * in either direction.
745 */
ceph_set_error_write(struct ceph_inode_info * ci)746 static inline void ceph_set_error_write(struct ceph_inode_info *ci)
747 {
748 set_bit(CEPH_I_ERROR_WRITE_BIT, &ci->i_ceph_flags);
749 }
750
ceph_clear_error_write(struct ceph_inode_info * ci)751 static inline void ceph_clear_error_write(struct ceph_inode_info *ci)
752 {
753 clear_bit(CEPH_I_ERROR_WRITE_BIT, &ci->i_ceph_flags);
754 }
755
__ceph_dir_set_complete(struct ceph_inode_info * ci,long long release_count,long long ordered_count)756 static inline void __ceph_dir_set_complete(struct ceph_inode_info *ci,
757 long long release_count,
758 long long ordered_count)
759 {
760 /*
761 * Makes sure operations that setup readdir cache (update page
762 * cache and i_size) are strongly ordered w.r.t. the following
763 * atomic64_set() operations.
764 */
765 smp_mb();
766 atomic64_set(&ci->i_complete_seq[0], release_count);
767 atomic64_set(&ci->i_complete_seq[1], ordered_count);
768 }
769
__ceph_dir_clear_complete(struct ceph_inode_info * ci)770 static inline void __ceph_dir_clear_complete(struct ceph_inode_info *ci)
771 {
772 atomic64_inc(&ci->i_release_count);
773 }
774
__ceph_dir_clear_ordered(struct ceph_inode_info * ci)775 static inline void __ceph_dir_clear_ordered(struct ceph_inode_info *ci)
776 {
777 atomic64_inc(&ci->i_ordered_count);
778 }
779
__ceph_dir_is_complete(struct ceph_inode_info * ci)780 static inline bool __ceph_dir_is_complete(struct ceph_inode_info *ci)
781 {
782 return atomic64_read(&ci->i_complete_seq[0]) ==
783 atomic64_read(&ci->i_release_count);
784 }
785
__ceph_dir_is_complete_ordered(struct ceph_inode_info * ci)786 static inline bool __ceph_dir_is_complete_ordered(struct ceph_inode_info *ci)
787 {
788 return atomic64_read(&ci->i_complete_seq[0]) ==
789 atomic64_read(&ci->i_release_count) &&
790 atomic64_read(&ci->i_complete_seq[1]) ==
791 atomic64_read(&ci->i_ordered_count);
792 }
793
ceph_dir_clear_complete(struct inode * inode)794 static inline void ceph_dir_clear_complete(struct inode *inode)
795 {
796 __ceph_dir_clear_complete(ceph_inode(inode));
797 }
798
ceph_dir_clear_ordered(struct inode * inode)799 static inline void ceph_dir_clear_ordered(struct inode *inode)
800 {
801 __ceph_dir_clear_ordered(ceph_inode(inode));
802 }
803
ceph_dir_is_complete_ordered(struct inode * inode)804 static inline bool ceph_dir_is_complete_ordered(struct inode *inode)
805 {
806 bool ret = __ceph_dir_is_complete_ordered(ceph_inode(inode));
807 smp_rmb();
808 return ret;
809 }
810
811 /* find a specific frag @f */
812 extern struct ceph_inode_frag *__ceph_find_frag(struct ceph_inode_info *ci,
813 u32 f);
814
815 /*
816 * choose fragment for value @v. copy frag content to pfrag, if leaf
817 * exists
818 */
819 extern u32 ceph_choose_frag(struct ceph_inode_info *ci, u32 v,
820 struct ceph_inode_frag *pfrag,
821 int *found);
822
ceph_dentry(const struct dentry * dentry)823 static inline struct ceph_dentry_info *ceph_dentry(const struct dentry *dentry)
824 {
825 return (struct ceph_dentry_info *)dentry->d_fsdata;
826 }
827
828 /*
829 * caps helpers
830 */
__ceph_is_any_real_caps(struct ceph_inode_info * ci)831 static inline bool __ceph_is_any_real_caps(struct ceph_inode_info *ci)
832 {
833 return !RB_EMPTY_ROOT(&ci->i_caps);
834 }
835
836 extern int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented);
837 extern int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int t);
838 extern int __ceph_caps_issued_mask_metric(struct ceph_inode_info *ci, int mask,
839 int t);
840 extern int __ceph_caps_issued_other(struct ceph_inode_info *ci,
841 struct ceph_cap *cap);
842
ceph_caps_issued(struct ceph_inode_info * ci)843 static inline int ceph_caps_issued(struct ceph_inode_info *ci)
844 {
845 int issued;
846 spin_lock(&ci->i_ceph_lock);
847 issued = __ceph_caps_issued(ci, NULL);
848 spin_unlock(&ci->i_ceph_lock);
849 return issued;
850 }
851
ceph_caps_issued_mask_metric(struct ceph_inode_info * ci,int mask,int touch)852 static inline int ceph_caps_issued_mask_metric(struct ceph_inode_info *ci,
853 int mask, int touch)
854 {
855 int r;
856 spin_lock(&ci->i_ceph_lock);
857 r = __ceph_caps_issued_mask_metric(ci, mask, touch);
858 spin_unlock(&ci->i_ceph_lock);
859 return r;
860 }
861
__ceph_caps_dirty(struct ceph_inode_info * ci)862 static inline int __ceph_caps_dirty(struct ceph_inode_info *ci)
863 {
864 return ci->i_dirty_caps | ci->i_flushing_caps;
865 }
866 extern struct ceph_cap_flush *ceph_alloc_cap_flush(void);
867 extern void ceph_free_cap_flush(struct ceph_cap_flush *cf);
868 extern int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask,
869 struct ceph_cap_flush **pcf);
870
871 extern int __ceph_caps_revoking_other(struct ceph_inode_info *ci,
872 struct ceph_cap *ocap, int mask);
873 extern int __ceph_caps_used(struct ceph_inode_info *ci);
874
__ceph_is_file_opened(struct ceph_inode_info * ci)875 static inline bool __ceph_is_file_opened(struct ceph_inode_info *ci)
876 {
877 return ci->i_nr_by_mode[0];
878 }
879 extern int __ceph_caps_file_wanted(struct ceph_inode_info *ci);
880 extern int __ceph_caps_wanted(struct ceph_inode_info *ci);
881
882 /* what the mds thinks we want */
883 extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci, bool check);
884
885 extern void ceph_caps_init(struct ceph_mds_client *mdsc);
886 extern void ceph_caps_finalize(struct ceph_mds_client *mdsc);
887 extern void ceph_adjust_caps_max_min(struct ceph_mds_client *mdsc,
888 struct ceph_mount_options *fsopt);
889 extern int ceph_reserve_caps(struct ceph_mds_client *mdsc,
890 struct ceph_cap_reservation *ctx, int need);
891 extern void ceph_unreserve_caps(struct ceph_mds_client *mdsc,
892 struct ceph_cap_reservation *ctx);
893 extern void ceph_reservation_status(struct ceph_fs_client *client,
894 int *total, int *avail, int *used,
895 int *reserved, int *min);
896 extern void change_auth_cap_ses(struct ceph_inode_info *ci,
897 struct ceph_mds_session *session);
898
899
900
901 /*
902 * we keep buffered readdir results attached to file->private_data
903 */
904 #define CEPH_F_SYNC 1
905 #define CEPH_F_ATEND 2
906
907 struct ceph_file_info {
908 short fmode; /* initialized on open */
909 short flags; /* CEPH_F_* */
910
911 spinlock_t rw_contexts_lock;
912 struct list_head rw_contexts;
913
914 u32 filp_gen;
915 };
916
917 struct ceph_dir_file_info {
918 struct ceph_file_info file_info;
919
920 /* readdir: position within the dir */
921 u32 frag;
922 struct ceph_mds_request *last_readdir;
923
924 /* readdir: position within a frag */
925 unsigned next_offset; /* offset of next chunk (last_name's + 1) */
926 char *last_name; /* last entry in previous chunk */
927 long long dir_release_count;
928 long long dir_ordered_count;
929 int readdir_cache_idx;
930
931 /* used for -o dirstat read() on directory thing */
932 char *dir_info;
933 int dir_info_len;
934 };
935
936 struct ceph_rw_context {
937 struct list_head list;
938 struct task_struct *thread;
939 int caps;
940 };
941
942 #define CEPH_DEFINE_RW_CONTEXT(_name, _caps) \
943 struct ceph_rw_context _name = { \
944 .thread = current, \
945 .caps = _caps, \
946 }
947
ceph_add_rw_context(struct ceph_file_info * cf,struct ceph_rw_context * ctx)948 static inline void ceph_add_rw_context(struct ceph_file_info *cf,
949 struct ceph_rw_context *ctx)
950 {
951 spin_lock(&cf->rw_contexts_lock);
952 list_add(&ctx->list, &cf->rw_contexts);
953 spin_unlock(&cf->rw_contexts_lock);
954 }
955
ceph_del_rw_context(struct ceph_file_info * cf,struct ceph_rw_context * ctx)956 static inline void ceph_del_rw_context(struct ceph_file_info *cf,
957 struct ceph_rw_context *ctx)
958 {
959 spin_lock(&cf->rw_contexts_lock);
960 list_del(&ctx->list);
961 spin_unlock(&cf->rw_contexts_lock);
962 }
963
964 static inline struct ceph_rw_context*
ceph_find_rw_context(struct ceph_file_info * cf)965 ceph_find_rw_context(struct ceph_file_info *cf)
966 {
967 struct ceph_rw_context *ctx, *found = NULL;
968 spin_lock(&cf->rw_contexts_lock);
969 list_for_each_entry(ctx, &cf->rw_contexts, list) {
970 if (ctx->thread == current) {
971 found = ctx;
972 break;
973 }
974 }
975 spin_unlock(&cf->rw_contexts_lock);
976 return found;
977 }
978
979 struct ceph_readdir_cache_control {
980 struct folio *folio;
981 struct dentry **dentries;
982 int index;
983 };
984
985 /*
986 * A "snap realm" describes a subset of the file hierarchy sharing
987 * the same set of snapshots that apply to it. The realms themselves
988 * are organized into a hierarchy, such that children inherit (some of)
989 * the snapshots of their parents.
990 *
991 * All inodes within the realm that have capabilities are linked into a
992 * per-realm list.
993 */
994 struct ceph_snap_realm {
995 u64 ino;
996 struct inode *inode;
997 atomic_t nref;
998 struct rb_node node;
999
1000 u64 created, seq;
1001 u64 parent_ino;
1002 u64 parent_since; /* snapid when our current parent became so */
1003
1004 u64 *prior_parent_snaps; /* snaps inherited from any parents we */
1005 u32 num_prior_parent_snaps; /* had prior to parent_since */
1006 u64 *snaps; /* snaps specific to this realm */
1007 u32 num_snaps;
1008
1009 struct ceph_snap_realm *parent;
1010 struct list_head children; /* list of child realms */
1011 struct list_head child_item;
1012
1013 struct list_head empty_item; /* if i have ref==0 */
1014
1015 struct list_head dirty_item; /* if realm needs new context */
1016
1017 struct list_head rebuild_item; /* rebuild snap realms _downward_ in hierarchy */
1018
1019 /* the current set of snaps for this realm */
1020 struct ceph_snap_context *cached_context;
1021
1022 struct list_head inodes_with_caps;
1023 spinlock_t inodes_with_caps_lock;
1024 };
1025
default_congestion_kb(void)1026 static inline int default_congestion_kb(void)
1027 {
1028 int congestion_kb;
1029
1030 /*
1031 * Copied from NFS
1032 *
1033 * congestion size, scale with available memory.
1034 *
1035 * 64MB: 8192k
1036 * 128MB: 11585k
1037 * 256MB: 16384k
1038 * 512MB: 23170k
1039 * 1GB: 32768k
1040 * 2GB: 46340k
1041 * 4GB: 65536k
1042 * 8GB: 92681k
1043 * 16GB: 131072k
1044 *
1045 * This allows larger machines to have larger/more transfers.
1046 * Limit the default to 256M
1047 */
1048 congestion_kb = (16*int_sqrt(totalram_pages())) << (PAGE_SHIFT-10);
1049 if (congestion_kb > 256*1024)
1050 congestion_kb = 256*1024;
1051
1052 return congestion_kb;
1053 }
1054
1055
1056 /* super.c */
1057 extern int ceph_force_reconnect(struct super_block *sb);
1058 /* snap.c */
1059 struct ceph_snap_realm *ceph_lookup_snap_realm(struct ceph_mds_client *mdsc,
1060 u64 ino);
1061 extern void ceph_get_snap_realm(struct ceph_mds_client *mdsc,
1062 struct ceph_snap_realm *realm);
1063 extern void ceph_put_snap_realm(struct ceph_mds_client *mdsc,
1064 struct ceph_snap_realm *realm);
1065 extern int ceph_update_snap_trace(struct ceph_mds_client *m,
1066 void *p, void *e, bool deletion,
1067 struct ceph_snap_realm **realm_ret);
1068 void ceph_change_snap_realm(struct inode *inode, struct ceph_snap_realm *realm);
1069 extern void ceph_handle_snap(struct ceph_mds_client *mdsc,
1070 struct ceph_mds_session *session,
1071 struct ceph_msg *msg);
1072 extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci,
1073 struct ceph_cap_snap *capsnap);
1074 extern void ceph_cleanup_global_and_empty_realms(struct ceph_mds_client *mdsc);
1075
1076 extern struct ceph_snapid_map *ceph_get_snapid_map(struct ceph_mds_client *mdsc,
1077 u64 snap);
1078 extern void ceph_put_snapid_map(struct ceph_mds_client* mdsc,
1079 struct ceph_snapid_map *sm);
1080 extern void ceph_trim_snapid_map(struct ceph_mds_client *mdsc);
1081 extern void ceph_cleanup_snapid_map(struct ceph_mds_client *mdsc);
1082 void ceph_umount_begin(struct super_block *sb);
1083
1084
1085 /*
1086 * a cap_snap is "pending" if it is still awaiting an in-progress
1087 * sync write (that may/may not still update size, mtime, etc.).
1088 */
__ceph_have_pending_cap_snap(struct ceph_inode_info * ci)1089 static inline bool __ceph_have_pending_cap_snap(struct ceph_inode_info *ci)
1090 {
1091 return !list_empty(&ci->i_cap_snaps) &&
1092 list_last_entry(&ci->i_cap_snaps, struct ceph_cap_snap,
1093 ci_item)->writing;
1094 }
1095
1096 /* inode.c */
1097 struct ceph_mds_reply_info_in;
1098 struct ceph_mds_reply_dirfrag;
1099 struct ceph_acl_sec_ctx;
1100
1101 extern const struct inode_operations ceph_file_iops;
1102
1103 extern struct inode *ceph_alloc_inode(struct super_block *sb);
1104 extern void ceph_evict_inode(struct inode *inode);
1105 extern void ceph_free_inode(struct inode *inode);
1106
1107 struct inode *ceph_new_inode(struct inode *dir, struct dentry *dentry,
1108 umode_t *mode, struct ceph_acl_sec_ctx *as_ctx);
1109 void ceph_as_ctx_to_req(struct ceph_mds_request *req,
1110 struct ceph_acl_sec_ctx *as_ctx);
1111
1112 extern struct inode *ceph_get_inode(struct super_block *sb,
1113 struct ceph_vino vino,
1114 struct inode *newino);
1115 extern struct inode *ceph_get_snapdir(struct inode *parent);
1116 extern int ceph_fill_file_size(struct inode *inode, int issued,
1117 u32 truncate_seq, u64 truncate_size, u64 size);
1118 extern void ceph_inode_set_subvolume(struct inode *inode, u64 subvolume_id);
1119 extern void ceph_fill_file_time(struct inode *inode, int issued,
1120 u64 time_warp_seq, struct timespec64 *ctime,
1121 struct timespec64 *mtime,
1122 struct timespec64 *atime);
1123 extern int ceph_fill_inode(struct inode *inode, struct page *locked_page,
1124 struct ceph_mds_reply_info_in *iinfo,
1125 struct ceph_mds_reply_dirfrag *dirinfo,
1126 struct ceph_mds_session *session, int cap_fmode,
1127 struct ceph_cap_reservation *caps_reservation);
1128 extern int ceph_fill_trace(struct super_block *sb,
1129 struct ceph_mds_request *req);
1130 extern int ceph_readdir_prepopulate(struct ceph_mds_request *req,
1131 struct ceph_mds_session *session);
1132
1133 extern bool ceph_inode_set_size(struct inode *inode, loff_t size);
1134 extern void __ceph_do_pending_vmtruncate(struct inode *inode);
1135
1136 void ceph_queue_inode_work(struct inode *inode, int work_bit);
1137
ceph_queue_vmtruncate(struct inode * inode)1138 static inline void ceph_queue_vmtruncate(struct inode *inode)
1139 {
1140 ceph_queue_inode_work(inode, CEPH_I_WORK_VMTRUNCATE);
1141 }
1142
ceph_queue_invalidate(struct inode * inode)1143 static inline void ceph_queue_invalidate(struct inode *inode)
1144 {
1145 ceph_queue_inode_work(inode, CEPH_I_WORK_INVALIDATE_PAGES);
1146 }
1147
ceph_queue_writeback(struct inode * inode)1148 static inline void ceph_queue_writeback(struct inode *inode)
1149 {
1150 ceph_queue_inode_work(inode, CEPH_I_WORK_WRITEBACK);
1151 }
1152
ceph_queue_check_caps(struct inode * inode)1153 static inline void ceph_queue_check_caps(struct inode *inode)
1154 {
1155 ceph_queue_inode_work(inode, CEPH_I_WORK_CHECK_CAPS);
1156 }
1157
ceph_queue_flush_snaps(struct inode * inode)1158 static inline void ceph_queue_flush_snaps(struct inode *inode)
1159 {
1160 ceph_queue_inode_work(inode, CEPH_I_WORK_FLUSH_SNAPS);
1161 }
1162
1163 extern int ceph_try_to_choose_auth_mds(struct inode *inode, int mask);
1164 extern int __ceph_do_getattr(struct inode *inode, struct page *locked_page,
1165 int mask, bool force);
ceph_do_getattr(struct inode * inode,int mask,bool force)1166 static inline int ceph_do_getattr(struct inode *inode, int mask, bool force)
1167 {
1168 return __ceph_do_getattr(inode, NULL, mask, force);
1169 }
1170 extern int ceph_permission(const struct mnt_idmap *idmap,
1171 struct inode *inode, int mask);
1172
1173 struct ceph_iattr {
1174 struct ceph_fscrypt_auth *fscrypt_auth;
1175 };
1176
1177 extern int __ceph_setattr(const struct mnt_idmap *idmap, struct inode *inode,
1178 struct iattr *attr, struct ceph_iattr *cia);
1179 extern int ceph_setattr(const struct mnt_idmap *idmap,
1180 struct dentry *dentry, struct iattr *attr);
1181 extern int ceph_getattr(const struct mnt_idmap *idmap,
1182 const struct path *path, struct kstat *stat,
1183 u32 request_mask, unsigned int flags);
1184 void ceph_inode_shutdown(struct inode *inode);
1185
ceph_inode_is_shutdown(struct inode * inode)1186 static inline bool ceph_inode_is_shutdown(struct inode *inode)
1187 {
1188 unsigned long flags = READ_ONCE(ceph_inode(inode)->i_ceph_flags);
1189 struct ceph_fs_client *fsc = ceph_inode_to_fs_client(inode);
1190 int state = READ_ONCE(fsc->mount_state);
1191
1192 return (flags & CEPH_I_SHUTDOWN) || state >= CEPH_MOUNT_SHUTDOWN;
1193 }
1194
1195 /* xattr.c */
1196 int __ceph_setxattr(struct inode *, const char *, const void *, size_t, int);
1197 int ceph_do_getvxattr(struct inode *inode, const char *name, void *value, size_t size);
1198 ssize_t __ceph_getxattr(struct inode *, const char *, void *, size_t);
1199 extern ssize_t ceph_listxattr(struct dentry *, char *, size_t);
1200 extern struct ceph_buffer *__ceph_build_xattrs_blob(struct ceph_inode_info *ci);
1201 extern void __ceph_destroy_xattrs(struct ceph_inode_info *ci);
1202 extern const struct xattr_handler * const ceph_xattr_handlers[];
1203
1204 struct ceph_acl_sec_ctx {
1205 #ifdef CONFIG_CEPH_FS_POSIX_ACL
1206 void *default_acl;
1207 void *acl;
1208 #endif
1209 #ifdef CONFIG_CEPH_FS_SECURITY_LABEL
1210 struct lsm_context lsmctx;
1211 #endif
1212 #ifdef CONFIG_FS_ENCRYPTION
1213 struct ceph_fscrypt_auth *fscrypt_auth;
1214 #endif
1215 struct ceph_pagelist *pagelist;
1216 };
1217
1218 #ifdef CONFIG_SECURITY
1219 extern bool ceph_security_xattr_deadlock(struct inode *in);
1220 extern bool ceph_security_xattr_wanted(struct inode *in);
1221 #else
ceph_security_xattr_deadlock(struct inode * in)1222 static inline bool ceph_security_xattr_deadlock(struct inode *in)
1223 {
1224 return false;
1225 }
ceph_security_xattr_wanted(struct inode * in)1226 static inline bool ceph_security_xattr_wanted(struct inode *in)
1227 {
1228 return false;
1229 }
1230 #endif
1231
1232 #ifdef CONFIG_CEPH_FS_SECURITY_LABEL
1233 extern int ceph_security_init_secctx(struct dentry *dentry, umode_t mode,
1234 struct ceph_acl_sec_ctx *ctx);
ceph_security_invalidate_secctx(struct inode * inode)1235 static inline void ceph_security_invalidate_secctx(struct inode *inode)
1236 {
1237 security_inode_invalidate_secctx(inode);
1238 }
1239 #else
ceph_security_init_secctx(struct dentry * dentry,umode_t mode,struct ceph_acl_sec_ctx * ctx)1240 static inline int ceph_security_init_secctx(struct dentry *dentry, umode_t mode,
1241 struct ceph_acl_sec_ctx *ctx)
1242 {
1243 return 0;
1244 }
ceph_security_invalidate_secctx(struct inode * inode)1245 static inline void ceph_security_invalidate_secctx(struct inode *inode)
1246 {
1247 }
1248 #endif
1249
1250 void ceph_release_acl_sec_ctx(struct ceph_acl_sec_ctx *as_ctx);
1251
1252 /* acl.c */
1253 #ifdef CONFIG_CEPH_FS_POSIX_ACL
1254
1255 struct posix_acl *ceph_get_acl(struct inode *, int, bool);
1256 int ceph_set_acl(const struct mnt_idmap *idmap,
1257 struct dentry *dentry, struct posix_acl *acl, int type);
1258 int ceph_pre_init_acls(struct inode *dir, umode_t *mode,
1259 struct ceph_acl_sec_ctx *as_ctx);
1260 void ceph_init_inode_acls(struct inode *inode,
1261 struct ceph_acl_sec_ctx *as_ctx);
1262
ceph_forget_all_cached_acls(struct inode * inode)1263 static inline void ceph_forget_all_cached_acls(struct inode *inode)
1264 {
1265 forget_all_cached_acls(inode);
1266 }
1267
1268 #else
1269
1270 #define ceph_get_acl NULL
1271 #define ceph_set_acl NULL
1272
ceph_pre_init_acls(struct inode * dir,umode_t * mode,struct ceph_acl_sec_ctx * as_ctx)1273 static inline int ceph_pre_init_acls(struct inode *dir, umode_t *mode,
1274 struct ceph_acl_sec_ctx *as_ctx)
1275 {
1276 return 0;
1277 }
ceph_init_inode_acls(struct inode * inode,struct ceph_acl_sec_ctx * as_ctx)1278 static inline void ceph_init_inode_acls(struct inode *inode,
1279 struct ceph_acl_sec_ctx *as_ctx)
1280 {
1281 }
1282
ceph_forget_all_cached_acls(struct inode * inode)1283 static inline void ceph_forget_all_cached_acls(struct inode *inode)
1284 {
1285 }
1286
1287 #endif
1288
1289 /* caps.c */
1290 extern const char *ceph_cap_string(int c);
1291 extern void ceph_handle_caps(struct ceph_mds_session *session,
1292 struct ceph_msg *msg);
1293 extern struct ceph_cap *ceph_get_cap(struct ceph_mds_client *mdsc,
1294 struct ceph_cap_reservation *ctx);
1295 extern void ceph_add_cap(struct inode *inode,
1296 struct ceph_mds_session *session, u64 cap_id,
1297 unsigned issued, unsigned wanted,
1298 unsigned cap, unsigned seq, u64 realmino, int flags,
1299 struct ceph_cap **new_cap);
1300
1301 /**
1302 * Determine whether __ceph_remove_cap() has been called on this #cap
1303 * (but the object has not yet been freed because it is protected by
1304 * `ceph_mds_session.s_cap_iterator`).
1305 *
1306 * Caller must lock either `ceph_inode_info.i_ceph_lock` or
1307 * `ceph_mds_session.s_cap_lock`.
1308 */
ceph_cap_is_removed(const struct ceph_cap * cap)1309 static inline bool ceph_cap_is_removed(const struct ceph_cap *cap)
1310 {
1311 return RB_EMPTY_NODE(&cap->ci_node);
1312 }
1313
1314 extern void ceph_remove_cap(struct ceph_mds_client *mdsc, struct ceph_cap *cap,
1315 struct ceph_inode_info *ci,
1316 bool queue_release);
1317 extern void __ceph_remove_caps(struct ceph_inode_info *ci);
1318 extern void ceph_put_cap(struct ceph_mds_client *mdsc,
1319 struct ceph_cap *cap);
1320 extern int ceph_is_any_caps(struct inode *inode);
1321
1322 extern int ceph_write_inode(struct inode *inode, struct writeback_control *wbc);
1323 extern int ceph_fsync(struct file *file, loff_t start, loff_t end,
1324 int datasync);
1325 extern void ceph_early_kick_flushing_caps(struct ceph_mds_client *mdsc,
1326 struct ceph_mds_session *session);
1327 extern void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc,
1328 struct ceph_mds_session *session);
1329 void ceph_kick_flushing_inode_caps(struct ceph_mds_session *session,
1330 struct ceph_inode_info *ci);
1331 extern struct ceph_cap *__get_cap_for_mds(struct ceph_inode_info *ci,
1332 int mds);
1333 extern struct ceph_cap *ceph_get_cap_for_mds(struct ceph_inode_info *ci,
1334 int mds);
1335 extern void ceph_take_cap_refs(struct ceph_inode_info *ci, int caps,
1336 bool snap_rwsem_locked);
1337 extern void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps);
1338 extern void ceph_put_cap_refs(struct ceph_inode_info *ci, int had);
1339 extern void ceph_put_cap_refs_async(struct ceph_inode_info *ci, int had);
1340 extern void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr,
1341 struct ceph_snap_context *snapc);
1342 extern void __ceph_remove_capsnap(struct inode *inode,
1343 struct ceph_cap_snap *capsnap,
1344 bool *wake_ci, bool *wake_mdsc);
1345 extern void ceph_remove_capsnap(struct inode *inode,
1346 struct ceph_cap_snap *capsnap,
1347 bool *wake_ci, bool *wake_mdsc);
1348 extern void ceph_flush_snaps(struct ceph_inode_info *ci,
1349 struct ceph_mds_session **psession);
1350 extern bool __ceph_should_report_size(struct ceph_inode_info *ci);
1351 extern void ceph_check_caps(struct ceph_inode_info *ci, int flags);
1352 extern unsigned long ceph_check_delayed_caps(struct ceph_mds_client *mdsc);
1353 extern void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc);
1354 extern void ceph_flush_cap_releases(struct ceph_mds_client *mdsc);
1355 extern int ceph_drop_caps_for_unlink(struct inode *inode);
1356 extern int ceph_encode_inode_release(void **p, struct inode *inode,
1357 int mds, int drop, int unless, int force);
1358 extern int ceph_encode_dentry_release(void **p, struct dentry *dn,
1359 struct inode *dir,
1360 int mds, int drop, int unless);
1361
1362 extern int __ceph_get_caps(struct inode *inode, struct ceph_file_info *fi,
1363 int need, int want, loff_t endoff, int *got);
1364 extern int ceph_get_caps(struct file *filp, int need, int want,
1365 loff_t endoff, int *got);
1366 extern int ceph_try_get_caps(struct inode *inode,
1367 int need, int want, bool nonblock, int *got);
1368
1369 /* for counting open files by mode */
1370 extern void ceph_get_fmode(struct ceph_inode_info *ci, int mode, int count);
1371 extern void ceph_put_fmode(struct ceph_inode_info *ci, int mode, int count);
1372 extern void __ceph_touch_fmode(struct ceph_inode_info *ci,
1373 struct ceph_mds_client *mdsc, int fmode);
1374
1375 /* addr.c */
1376 extern const struct address_space_operations ceph_aops;
1377 extern const struct netfs_request_ops ceph_netfs_ops;
1378 int ceph_mmap_prepare(struct vm_area_desc *desc);
1379 extern int ceph_uninline_data(struct file *file);
1380 extern int ceph_pool_perm_check(struct inode *inode, int need);
1381 extern void ceph_pool_perm_destroy(struct ceph_mds_client* mdsc);
1382 int ceph_purge_inode_cap(struct inode *inode, struct ceph_cap *cap, bool *invalidate);
1383
ceph_has_inline_data(struct ceph_inode_info * ci)1384 static inline bool ceph_has_inline_data(struct ceph_inode_info *ci)
1385 {
1386 if (ci->i_inline_version == CEPH_INLINE_NONE ||
1387 ci->i_inline_version == 1) /* initial version, no data */
1388 return false;
1389 return true;
1390 }
1391
1392 /* file.c */
1393 extern const struct file_operations ceph_file_fops;
1394
1395 extern int ceph_renew_caps(struct inode *inode, int fmode);
1396 extern int ceph_open(struct inode *inode, struct file *file);
1397 extern int ceph_atomic_open(struct inode *dir, struct dentry *dentry,
1398 struct file *file, unsigned flags, umode_t mode);
1399 extern ssize_t __ceph_sync_read(struct inode *inode, loff_t *ki_pos,
1400 struct iov_iter *to, int *retry_op,
1401 u64 *last_objver);
1402 extern int ceph_release(struct inode *inode, struct file *filp);
1403 extern void ceph_fill_inline_data(struct inode *inode, struct page *locked_page,
1404 char *data, size_t len);
1405
1406 /* dir.c */
1407 extern const struct file_operations ceph_dir_fops;
1408 extern const struct file_operations ceph_snapdir_fops;
1409 extern const struct inode_operations ceph_dir_iops;
1410 extern const struct inode_operations ceph_snapdir_iops;
1411 extern const struct dentry_operations ceph_dentry_ops;
1412
1413 extern loff_t ceph_make_fpos(unsigned high, unsigned off, bool hash_order);
1414 extern int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry);
1415 extern struct dentry *ceph_handle_snapdir(struct ceph_mds_request *req,
1416 struct dentry *dentry);
1417 extern struct dentry *ceph_finish_lookup(struct ceph_mds_request *req,
1418 struct dentry *dentry, int err);
1419
1420 extern void __ceph_dentry_lease_touch(struct ceph_dentry_info *di);
1421 extern void __ceph_dentry_dir_lease_touch(struct ceph_dentry_info *di);
1422 extern void ceph_invalidate_dentry_lease(struct dentry *dentry);
1423 extern int ceph_trim_dentries(struct ceph_mds_client *mdsc);
1424 extern unsigned ceph_dentry_hash(struct inode *dir, struct dentry *dn);
1425 extern void ceph_readdir_cache_release(struct ceph_readdir_cache_control *ctl);
1426
1427 /* ioctl.c */
1428 extern long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
1429
1430 /* export.c */
1431 extern const struct export_operations ceph_export_ops;
1432 struct inode *ceph_lookup_inode(struct super_block *sb, u64 ino);
1433
1434 /* locks.c */
1435 extern __init void ceph_flock_init(void);
1436 extern int ceph_lock(struct file *file, int cmd, struct file_lock *fl);
1437 extern int ceph_flock(struct file *file, int cmd, struct file_lock *fl);
1438 extern void ceph_count_locks(struct inode *inode, int *p_num, int *f_num);
1439 extern int ceph_encode_locks_to_buffer(struct inode *inode,
1440 struct ceph_filelock *flocks,
1441 int num_fcntl_locks,
1442 int num_flock_locks);
1443 extern int ceph_locks_to_pagelist(struct ceph_filelock *flocks,
1444 struct ceph_pagelist *pagelist,
1445 int num_fcntl_locks, int num_flock_locks);
1446
1447 /* debugfs.c */
1448 extern void ceph_fs_debugfs_init(struct ceph_fs_client *client);
1449 extern void ceph_fs_debugfs_cleanup(struct ceph_fs_client *client);
1450
1451 /* quota.c */
1452
1453 enum quota_get_realm {
1454 QUOTA_GET_MAX_FILES,
1455 QUOTA_GET_MAX_BYTES,
1456 QUOTA_GET_ANY
1457 };
1458
__ceph_has_quota(struct ceph_inode_info * ci,enum quota_get_realm which)1459 static inline bool __ceph_has_quota(struct ceph_inode_info *ci,
1460 enum quota_get_realm which)
1461 {
1462 bool has_quota = false;
1463
1464 switch (which) {
1465 case QUOTA_GET_MAX_BYTES:
1466 has_quota = !!ci->i_max_bytes;
1467 break;
1468 case QUOTA_GET_MAX_FILES:
1469 has_quota = !!ci->i_max_files;
1470 break;
1471 default:
1472 has_quota = !!(ci->i_max_files || ci->i_max_bytes);
1473 }
1474 return has_quota;
1475 }
1476
1477 extern void ceph_adjust_quota_realms_count(struct inode *inode, bool inc);
1478
__ceph_update_quota(struct ceph_inode_info * ci,u64 max_bytes,u64 max_files)1479 static inline void __ceph_update_quota(struct ceph_inode_info *ci,
1480 u64 max_bytes, u64 max_files)
1481 {
1482 bool had_quota, has_quota;
1483 had_quota = __ceph_has_quota(ci, QUOTA_GET_ANY);
1484 ci->i_max_bytes = max_bytes;
1485 ci->i_max_files = max_files;
1486 has_quota = __ceph_has_quota(ci, QUOTA_GET_ANY);
1487
1488 if (had_quota != has_quota)
1489 ceph_adjust_quota_realms_count(&ci->netfs.inode, has_quota);
1490 }
1491
__ceph_sparse_read_ext_count(struct inode * inode,u64 len)1492 static inline int __ceph_sparse_read_ext_count(struct inode *inode, u64 len)
1493 {
1494 int cnt = 0;
1495
1496 if (IS_ENCRYPTED(inode)) {
1497 cnt = len >> CEPH_FSCRYPT_BLOCK_SHIFT;
1498 if (cnt > CEPH_SPARSE_EXT_ARRAY_INITIAL)
1499 cnt = 0;
1500 }
1501
1502 return cnt;
1503 }
1504
1505 extern void ceph_handle_quota(struct ceph_mds_client *mdsc,
1506 struct ceph_mds_session *session,
1507 struct ceph_msg *msg);
1508 extern bool ceph_quota_is_max_files_exceeded(struct inode *inode);
1509 extern bool ceph_quota_is_same_realm(struct inode *old, struct inode *new);
1510 extern bool ceph_quota_is_max_bytes_exceeded(struct inode *inode,
1511 loff_t newlen);
1512 extern bool ceph_quota_is_max_bytes_approaching(struct inode *inode,
1513 loff_t newlen);
1514 extern bool ceph_quota_update_statfs(struct ceph_fs_client *fsc,
1515 struct kstatfs *buf);
1516 extern void ceph_cleanup_quotarealms_inodes(struct ceph_mds_client *mdsc);
1517
1518 bool ceph_inc_mds_stopping_blocker(struct ceph_mds_client *mdsc,
1519 struct ceph_mds_session *session);
1520 void ceph_dec_mds_stopping_blocker(struct ceph_mds_client *mdsc);
1521 bool ceph_inc_osd_stopping_blocker(struct ceph_mds_client *mdsc);
1522 void ceph_dec_osd_stopping_blocker(struct ceph_mds_client *mdsc);
1523 #endif /* _FS_CEPH_SUPER_H */
1524