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