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