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