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