xref: /linux/fs/ceph/super.h (revision ac2dc6d57425ffa9629941d7c9d7c0e51082cb5a)
1 /* SPDX-License-Identifier: GPL-2.0 */
2 #ifndef _FS_CEPH_SUPER_H
3 #define _FS_CEPH_SUPER_H
4 
5 #include <linux/ceph/ceph_debug.h>
6 #include <linux/ceph/osd_client.h>
7 
8 #include <linux/unaligned.h>
9 #include <linux/backing-dev.h>
10 #include <linux/completion.h>
11 #include <linux/exportfs.h>
12 #include <linux/fs.h>
13 #include <linux/mempool.h>
14 #include <linux/pagemap.h>
15 #include <linux/wait.h>
16 #include <linux/writeback.h>
17 #include <linux/slab.h>
18 #include <linux/posix_acl.h>
19 #include <linux/refcount.h>
20 #include <linux/security.h>
21 #include <linux/netfs.h>
22 #include <linux/fscache.h>
23 #include <linux/hashtable.h>
24 
25 #include <linux/ceph/libceph.h>
26 #include "crypto.h"
27 
28 /* large granularity for statfs utilization stats to facilitate
29  * large volume sizes on 32-bit machines. */
30 #define CEPH_BLOCK_SHIFT   22  /* 4 MB */
31 #define CEPH_BLOCK         (1 << CEPH_BLOCK_SHIFT)
32 #define CEPH_4K_BLOCK_SHIFT 12  /* 4 KB */
33 
34 #define CEPH_MOUNT_OPT_CLEANRECOVER    (1<<1) /* auto reonnect (clean mode) after blocklisted */
35 #define CEPH_MOUNT_OPT_DIRSTAT         (1<<4) /* `cat dirname` for stats */
36 #define CEPH_MOUNT_OPT_RBYTES          (1<<5) /* dir st_bytes = rbytes */
37 #define CEPH_MOUNT_OPT_NOASYNCREADDIR  (1<<7) /* no dcache readdir */
38 #define CEPH_MOUNT_OPT_INO32           (1<<8) /* 32 bit inos */
39 #define CEPH_MOUNT_OPT_DCACHE          (1<<9) /* use dcache for readdir etc */
40 #define CEPH_MOUNT_OPT_FSCACHE         (1<<10) /* use fscache */
41 #define CEPH_MOUNT_OPT_NOPOOLPERM      (1<<11) /* no pool permission check */
42 #define CEPH_MOUNT_OPT_MOUNTWAIT       (1<<12) /* mount waits if no mds is up */
43 #define CEPH_MOUNT_OPT_NOQUOTADF       (1<<13) /* no root dir quota in statfs */
44 #define CEPH_MOUNT_OPT_NOCOPYFROM      (1<<14) /* don't use RADOS 'copy-from' op */
45 #define CEPH_MOUNT_OPT_ASYNC_DIROPS    (1<<15) /* allow async directory ops */
46 #define CEPH_MOUNT_OPT_NOPAGECACHE     (1<<16) /* bypass pagecache altogether */
47 #define CEPH_MOUNT_OPT_SPARSEREAD      (1<<17) /* always do sparse reads */
48 
49 #define CEPH_MOUNT_OPT_DEFAULT			\
50 	(CEPH_MOUNT_OPT_DCACHE |		\
51 	 CEPH_MOUNT_OPT_NOCOPYFROM |		\
52 	 CEPH_MOUNT_OPT_ASYNC_DIROPS)
53 
54 #define ceph_set_mount_opt(fsc, opt) \
55 	(fsc)->mount_options->flags |= CEPH_MOUNT_OPT_##opt
56 #define ceph_clear_mount_opt(fsc, opt) \
57 	(fsc)->mount_options->flags &= ~CEPH_MOUNT_OPT_##opt
58 #define ceph_test_mount_opt(fsc, opt) \
59 	(!!((fsc)->mount_options->flags & CEPH_MOUNT_OPT_##opt))
60 
61 /* max size of osd read request, limited by libceph */
62 #define CEPH_MAX_READ_SIZE              CEPH_MSG_MAX_DATA_LEN
63 /* osd has a configurable limitation of max write size.
64  * CEPH_MSG_MAX_DATA_LEN should be small enough. */
65 #define CEPH_MAX_WRITE_SIZE		CEPH_MSG_MAX_DATA_LEN
66 #define CEPH_RASIZE_DEFAULT             (8192*1024)    /* max readahead */
67 #define CEPH_MAX_READDIR_DEFAULT        1024
68 #define CEPH_MAX_READDIR_BYTES_DEFAULT  (512*1024)
69 #define CEPH_SNAPDIRNAME_DEFAULT        ".snap"
70 
71 /*
72  * Delay telling the MDS we no longer want caps, in case we reopen
73  * the file.  Delay a minimum amount of time, even if we send a cap
74  * message for some other reason.  Otherwise, take the oppotunity to
75  * update the mds to avoid sending another message later.
76  */
77 #define CEPH_CAPS_WANTED_DELAY_MIN_DEFAULT      5  /* cap release delay */
78 #define CEPH_CAPS_WANTED_DELAY_MAX_DEFAULT     60  /* cap release delay */
79 
80 struct ceph_mount_options {
81 	unsigned int flags;
82 
83 	unsigned int wsize;            /* max write size */
84 	unsigned int rsize;            /* max read size */
85 	unsigned int rasize;           /* max readahead */
86 	unsigned int congestion_kb;    /* max writeback in flight */
87 	unsigned int caps_wanted_delay_min, caps_wanted_delay_max;
88 	int caps_max;
89 	unsigned int max_readdir;       /* max readdir result (entries) */
90 	unsigned int max_readdir_bytes; /* max readdir result (bytes) */
91 
92 	bool new_dev_syntax;
93 
94 	/*
95 	 * everything above this point can be memcmp'd; everything below
96 	 * is handled in compare_mount_options()
97 	 */
98 
99 	char *snapdir_name;   /* default ".snap" */
100 	char *mds_namespace;  /* default NULL */
101 	char *server_path;    /* default NULL (means "/") */
102 	char *fscache_uniq;   /* default NULL */
103 	char *mon_addr;
104 	struct fscrypt_dummy_policy dummy_enc_policy;
105 };
106 
107 #define CEPH_NAMESPACE_WILDCARD		"*"
108 
ceph_namespace_match(const char * pattern,const char * target)109 static inline bool ceph_namespace_match(const char *pattern,
110 					const char *target)
111 {
112 	if (!pattern || !pattern[0] ||
113 	    !strcmp(pattern, CEPH_NAMESPACE_WILDCARD))
114 		return true;
115 
116 	return !strcmp(pattern, target);
117 }
118 
119 /*
120  * Check if the mds namespace in ceph_mount_options matches
121  * the passed in namespace string. First time match (when
122  * ->mds_namespace is NULL) is treated specially, since
123  * ->mds_namespace needs to be initialized by the caller.
124  */
namespace_equals(struct ceph_mount_options * fsopt,const char * namespace,size_t len)125 static inline bool namespace_equals(struct ceph_mount_options *fsopt,
126 				    const char *namespace, size_t len)
127 {
128 	return !(fsopt->mds_namespace &&
129 		 (strlen(fsopt->mds_namespace) != len ||
130 		  strncmp(fsopt->mds_namespace, namespace, len)));
131 }
132 
133 /* mount state */
134 enum {
135 	CEPH_MOUNT_MOUNTING,
136 	CEPH_MOUNT_MOUNTED,
137 	CEPH_MOUNT_UNMOUNTING,
138 	CEPH_MOUNT_UNMOUNTED,
139 	CEPH_MOUNT_SHUTDOWN,
140 	CEPH_MOUNT_RECOVER,
141 	CEPH_MOUNT_FENCE_IO,
142 };
143 
144 #define CEPH_ASYNC_CREATE_CONFLICT_BITS 8
145 
146 struct ceph_fs_client {
147 	struct super_block *sb;
148 
149 	struct list_head metric_wakeup;
150 
151 	struct ceph_mount_options *mount_options;
152 	struct ceph_client *client;
153 
154 	int mount_state;
155 
156 	bool blocklisted;
157 
158 	bool have_copy_from2;
159 
160 	u32 filp_gen;
161 	loff_t max_file_size;
162 
163 	struct ceph_mds_client *mdsc;
164 
165 	atomic_long_t writeback_count;
166 	bool write_congested;
167 
168 	struct workqueue_struct *inode_wq;
169 	struct workqueue_struct *cap_wq;
170 
171 	DECLARE_HASHTABLE(async_unlink_conflict, CEPH_ASYNC_CREATE_CONFLICT_BITS);
172 	spinlock_t async_unlink_conflict_lock;
173 
174 #ifdef CONFIG_DEBUG_FS
175 	struct dentry *debugfs_dentry_lru, *debugfs_caps;
176 	struct dentry *debugfs_congestion_kb;
177 	struct dentry *debugfs_bdi;
178 	struct dentry *debugfs_mdsc, *debugfs_mdsmap;
179 	struct dentry *debugfs_status;
180 	struct dentry *debugfs_mds_sessions;
181 	struct dentry *debugfs_metrics_dir;
182 	struct dentry *debugfs_subvolume_metrics;
183 #endif
184 
185 #ifdef CONFIG_CEPH_FSCACHE
186 	struct fscache_volume *fscache;
187 #endif
188 #ifdef CONFIG_FS_ENCRYPTION
189 	struct fscrypt_dummy_policy fsc_dummy_enc_policy;
190 #endif
191 };
192 
193 /*
194  * File i/o capability.  This tracks shared state with the metadata
195  * server that allows us to cache or writeback attributes or to read
196  * and write data.  For any given inode, we should have one or more
197  * capabilities, one issued by each metadata server, and our
198  * cumulative access is the OR of all issued capabilities.
199  *
200  * Each cap is referenced by the inode's i_caps rbtree and by per-mds
201  * session capability lists.
202  */
203 struct ceph_cap {
204 	struct ceph_inode_info *ci;
205 	struct rb_node ci_node;          /* per-ci cap tree */
206 	struct ceph_mds_session *session;
207 	struct list_head session_caps;   /* per-session caplist */
208 	u64 cap_id;       /* unique cap id (mds provided) */
209 	union {
210 		/* in-use caps */
211 		struct {
212 			int issued;       /* latest, from the mds */
213 			int implemented;  /* implemented superset of
214 					     issued (for revocation) */
215 			int mds;	  /* mds index for this cap */
216 			int mds_wanted;   /* caps wanted from this mds */
217 		};
218 		/* caps to release */
219 		struct {
220 			u64 cap_ino;
221 			int queue_release;
222 		};
223 	};
224 	u32 seq, issue_seq, mseq;
225 	u32 cap_gen;      /* active/stale cycle */
226 	unsigned long last_used;
227 	struct list_head caps_item;
228 };
229 
230 #define CHECK_CAPS_AUTHONLY     1  /* only check auth cap */
231 #define CHECK_CAPS_FLUSH        2  /* flush any dirty caps */
232 #define CHECK_CAPS_NOINVAL      4  /* don't invalidate pagecache */
233 #define CHECK_CAPS_FLUSH_FORCE  8  /* force flush any caps */
234 
235 struct ceph_cap_flush {
236 	u64 tid;
237 	int caps;
238 	bool wake; /* wake up flush waiters when finish ? */
239 	bool is_capsnap; /* true means capsnap */
240 	struct list_head g_list; // global
241 	struct list_head i_list; // per inode
242 };
243 
244 /*
245  * Snapped cap state that is pending flush to mds.  When a snapshot occurs,
246  * we first complete any in-process sync writes and writeback any dirty
247  * data before flushing the snapped state (tracked here) back to the MDS.
248  */
249 struct ceph_cap_snap {
250 	refcount_t nref;
251 	struct list_head ci_item;
252 
253 	struct ceph_cap_flush cap_flush;
254 
255 	u64 follows;
256 	int issued, dirty;
257 	struct ceph_snap_context *context;
258 
259 	umode_t mode;
260 	kuid_t uid;
261 	kgid_t gid;
262 
263 	struct ceph_buffer *xattr_blob;
264 	u64 xattr_version;
265 
266 	u64 size;
267 	u64 change_attr;
268 	struct timespec64 mtime, atime, ctime, btime;
269 	u64 time_warp_seq;
270 	u64 truncate_size;
271 	u32 truncate_seq;
272 	int writing;   /* a sync write is still in progress */
273 	int dirty_pages;     /* dirty pages awaiting writeback */
274 	bool inline_data;
275 	bool need_flush;
276 };
277 
ceph_put_cap_snap(struct ceph_cap_snap * capsnap)278 static inline void ceph_put_cap_snap(struct ceph_cap_snap *capsnap)
279 {
280 	if (refcount_dec_and_test(&capsnap->nref)) {
281 		if (capsnap->xattr_blob)
282 			ceph_buffer_put(capsnap->xattr_blob);
283 		kmem_cache_free(ceph_cap_snap_cachep, capsnap);
284 	}
285 }
286 
287 /*
288  * The frag tree describes how a directory is fragmented, potentially across
289  * multiple metadata servers.  It is also used to indicate points where
290  * metadata authority is delegated, and whether/where metadata is replicated.
291  *
292  * A _leaf_ frag will be present in the i_fragtree IFF there is
293  * delegation info.  That is, if mds >= 0 || ndist > 0.
294  */
295 #define CEPH_MAX_DIRFRAG_REP 4
296 
297 struct ceph_inode_frag {
298 	struct rb_node node;
299 
300 	/* fragtree state */
301 	u32 frag;
302 	int split_by;         /* i.e. 2^(split_by) children */
303 
304 	/* delegation and replication info */
305 	int mds;              /* -1 if same authority as parent */
306 	int ndist;            /* >0 if replicated */
307 	int dist[CEPH_MAX_DIRFRAG_REP];
308 };
309 
310 /*
311  * We cache inode xattrs as an encoded blob until they are first used,
312  * at which point we parse them into an rbtree.
313  */
314 struct ceph_inode_xattr {
315 	struct rb_node node;
316 
317 	const char *name;
318 	int name_len;
319 	const char *val;
320 	int val_len;
321 	int dirty;
322 
323 	int should_free_name;
324 	int should_free_val;
325 };
326 
327 /*
328  * Ceph dentry state
329  */
330 struct ceph_dentry_info {
331 	struct dentry *dentry;
332 	struct ceph_mds_session *lease_session;
333 	struct list_head lease_list;
334 	struct hlist_node hnode;
335 	unsigned long flags;
336 	int lease_shared_gen;
337 	u32 lease_gen;
338 	u32 lease_seq;
339 	unsigned long lease_renew_after, lease_renew_from;
340 	unsigned long time;
341 	u64 offset;
342 };
343 
344 #define CEPH_DENTRY_REFERENCED		(1 << 0)
345 #define CEPH_DENTRY_LEASE_LIST		(1 << 1)
346 #define CEPH_DENTRY_SHRINK_LIST		(1 << 2)
347 #define CEPH_DENTRY_PRIMARY_LINK	(1 << 3)
348 #define CEPH_DENTRY_ASYNC_UNLINK_BIT	(4)
349 #define CEPH_DENTRY_ASYNC_UNLINK	(1 << CEPH_DENTRY_ASYNC_UNLINK_BIT)
350 #define CEPH_DENTRY_ASYNC_CREATE_BIT	(5)
351 #define CEPH_DENTRY_ASYNC_CREATE	(1 << CEPH_DENTRY_ASYNC_CREATE_BIT)
352 
353 struct ceph_inode_xattrs_info {
354 	/*
355 	 * (still encoded) xattr blob. we avoid the overhead of parsing
356 	 * this until someone actually calls getxattr, etc.
357 	 *
358 	 * blob->vec.iov_len == 4 implies there are no xattrs; blob ==
359 	 * NULL means we don't know.
360 	*/
361 	struct ceph_buffer *blob, *prealloc_blob;
362 
363 	struct rb_root index;
364 	bool dirty;
365 	int count;
366 	int names_size;
367 	int vals_size;
368 	u64 version, index_version;
369 };
370 
371 /*
372  * Ceph inode.
373  */
374 struct ceph_inode_info {
375 	struct netfs_inode netfs; /* Netfslib context and vfs inode */
376 	struct ceph_vino i_vino;   /* ceph ino + snap */
377 
378 	spinlock_t i_ceph_lock;
379 
380 	u64 i_version;
381 	u64 i_inline_version;
382 	u32 i_time_warp_seq;
383 
384 	unsigned long i_ceph_flags;
385 	atomic64_t i_release_count;
386 	atomic64_t i_ordered_count;
387 	atomic64_t i_complete_seq[2];
388 
389 	struct ceph_dir_layout i_dir_layout;
390 	struct ceph_file_layout i_layout;
391 	struct ceph_file_layout i_cached_layout;	// for async creates
392 	char *i_symlink;
393 
394 	/* for dirs */
395 	struct timespec64 i_rctime;
396 	u64 i_rbytes, i_rfiles, i_rsubdirs, i_rsnaps;
397 	u64 i_files, i_subdirs;
398 
399 	/* quotas */
400 	u64 i_max_bytes, i_max_files;
401 
402 	/*
403 	 * Subvolume ID this inode belongs to. CEPH_SUBVOLUME_ID_NONE (0)
404 	 * means unknown/unset, matching the FUSE client convention.
405 	 * Once set to a valid (non-zero) value, it should not change
406 	 * during the inode's lifetime.
407 	 */
408 #define CEPH_SUBVOLUME_ID_NONE 0
409 	u64 i_subvolume_id;
410 
411 	s32 i_dir_pin;
412 
413 	struct rb_root i_fragtree;
414 	int i_fragtree_nsplits;
415 	struct mutex i_fragtree_mutex;
416 
417 	struct ceph_inode_xattrs_info i_xattrs;
418 
419 	/* capabilities.  protected _both_ by i_ceph_lock and cap->session's
420 	 * s_mutex. */
421 	struct rb_root i_caps;           /* cap list */
422 	struct ceph_cap *i_auth_cap;     /* authoritative cap, if any */
423 	unsigned i_dirty_caps, i_flushing_caps;     /* mask of dirtied fields */
424 
425 	/*
426 	 * Link to the auth cap's session's s_cap_dirty list. s_cap_dirty
427 	 * is protected by the mdsc->cap_dirty_lock, but each individual item
428 	 * is also protected by the inode's i_ceph_lock. Walking s_cap_dirty
429 	 * requires the mdsc->cap_dirty_lock. List presence for an item can
430 	 * be tested under the i_ceph_lock. Changing anything requires both.
431 	 */
432 	struct list_head i_dirty_item;
433 
434 	/*
435 	 * Link to session's s_cap_flushing list. Protected in a similar
436 	 * fashion to i_dirty_item, but also by the s_mutex for changes. The
437 	 * s_cap_flushing list can be walked while holding either the s_mutex
438 	 * or msdc->cap_dirty_lock. List presence can also be checked while
439 	 * holding the i_ceph_lock for this inode.
440 	 */
441 	struct list_head i_flushing_item;
442 
443 	/* we need to track cap writeback on a per-cap-bit basis, to allow
444 	 * overlapping, pipelined cap flushes to the mds.  we can probably
445 	 * reduce the tid to 8 bits if we're concerned about inode size. */
446 	struct ceph_cap_flush *i_prealloc_cap_flush;
447 	struct list_head i_cap_flush_list;
448 	wait_queue_head_t i_cap_wq;      /* threads waiting on a capability */
449 	unsigned long i_hold_caps_max; /* jiffies */
450 	struct list_head i_cap_delay_list;  /* for delayed cap release to mds */
451 	struct ceph_cap_reservation i_cap_migration_resv;
452 	struct list_head i_cap_snaps;   /* snapped state pending flush to mds */
453 	struct ceph_snap_context *i_head_snapc;  /* set if wr_buffer_head > 0 or
454 						    dirty|flushing caps */
455 	unsigned i_snap_caps;           /* cap bits for snapped files */
456 
457 	unsigned long i_last_rd;
458 	unsigned long i_last_wr;
459 	int i_nr_by_mode[CEPH_FILE_MODE_BITS];  /* open file counts */
460 
461 	struct mutex i_truncate_mutex;
462 	u32 i_truncate_seq;        /* last truncate to smaller size */
463 	u64 i_truncate_size;       /*  and the size we last truncated down to */
464 	int i_truncate_pending;    /*  still need to call vmtruncate */
465 	/*
466 	 * For none fscrypt case it equals to i_truncate_size or it will
467 	 * equals to fscrypt_file_size
468 	 */
469 	u64 i_truncate_pagecache_size;
470 
471 	u64 i_max_size;            /* max file size authorized by mds */
472 	u64 i_reported_size; /* (max_)size reported to or requested of mds */
473 	u64 i_wanted_max_size;     /* offset we'd like to write too */
474 	u64 i_requested_max_size;  /* max_size we've requested */
475 
476 	/* held references to caps */
477 	int i_pin_ref;
478 	int i_rd_ref, i_rdcache_ref, i_wr_ref, i_wb_ref, i_fx_ref;
479 	int i_wrbuffer_ref, i_wrbuffer_ref_head;
480 	atomic_t i_filelock_ref;
481 	atomic_t i_shared_gen;       /* increment each time we get FILE_SHARED */
482 	u32 i_rdcache_gen;      /* incremented each time we get FILE_CACHE. */
483 	u32 i_rdcache_revoking; /* RDCACHE gen to async invalidate, if any */
484 
485 	struct list_head i_unsafe_dirops; /* uncommitted mds dir ops */
486 	struct list_head i_unsafe_iops;   /* uncommitted mds inode ops */
487 	spinlock_t i_unsafe_lock;
488 
489 	union {
490 		struct ceph_snap_realm *i_snap_realm; /* snap realm (if caps) */
491 		struct ceph_snapid_map *i_snapid_map; /* snapid -> dev_t */
492 	};
493 	struct list_head i_snap_realm_item;
494 	struct list_head i_snap_flush_item;
495 	struct timespec64 i_btime;
496 	struct timespec64 i_snap_btime;
497 
498 	struct work_struct i_work;
499 	unsigned long  i_work_mask;
500 
501 #ifdef CONFIG_FS_ENCRYPTION
502 	struct fscrypt_inode_info *i_crypt_info;
503 	u32 fscrypt_auth_len;
504 	u32 fscrypt_file_len;
505 	u8 *fscrypt_auth;
506 	u8 *fscrypt_file;
507 #endif
508 };
509 
510 struct ceph_netfs_request_data {
511 	int caps;
512 
513 	/*
514 	 * Maximum size of a file readahead request.
515 	 * The fadvise could update the bdi's default ra_pages.
516 	 */
517 	unsigned int file_ra_pages;
518 
519 	/* Set it if fadvise disables file readahead entirely */
520 	bool file_ra_disabled;
521 };
522 
523 static inline struct ceph_inode_info *
ceph_inode(const struct inode * inode)524 ceph_inode(const struct inode *inode)
525 {
526 	return container_of(inode, struct ceph_inode_info, netfs.inode);
527 }
528 
529 static inline struct ceph_fs_client *
ceph_inode_to_fs_client(const struct inode * inode)530 ceph_inode_to_fs_client(const struct inode *inode)
531 {
532 	return (struct ceph_fs_client *)inode->i_sb->s_fs_info;
533 }
534 
535 static inline struct ceph_fs_client *
ceph_sb_to_fs_client(const struct super_block * sb)536 ceph_sb_to_fs_client(const struct super_block *sb)
537 {
538 	return (struct ceph_fs_client *)sb->s_fs_info;
539 }
540 
541 static inline struct ceph_mds_client *
ceph_sb_to_mdsc(const struct super_block * sb)542 ceph_sb_to_mdsc(const struct super_block *sb)
543 {
544 	return (struct ceph_mds_client *)ceph_sb_to_fs_client(sb)->mdsc;
545 }
546 
547 static inline struct ceph_client *
ceph_inode_to_client(const struct inode * inode)548 ceph_inode_to_client(const struct inode *inode)
549 {
550 	return (struct ceph_client *)ceph_inode_to_fs_client(inode)->client;
551 }
552 
553 static inline struct ceph_vino
ceph_vino(const struct inode * inode)554 ceph_vino(const struct inode *inode)
555 {
556 	return ceph_inode(inode)->i_vino;
557 }
558 
ceph_ino_to_ino32(u64 vino)559 static inline u32 ceph_ino_to_ino32(u64 vino)
560 {
561 	u32 ino = vino & 0xffffffff;
562 	ino ^= vino >> 32;
563 	if (!ino)
564 		ino = 2;
565 	return ino;
566 }
567 
568 /*
569  * Inode numbers in cephfs are 64 bits, but inode->i_ino is 32-bits on
570  * some arches. We generally do not use this value inside the ceph driver, but
571  * we do want to set it to something, so that generic vfs code has an
572  * appropriate value for tracepoints and the like.
573  */
ceph_vino_to_ino_t(struct ceph_vino vino)574 static inline ino_t ceph_vino_to_ino_t(struct ceph_vino vino)
575 {
576 	if (sizeof(ino_t) == sizeof(u32))
577 		return ceph_ino_to_ino32(vino.ino);
578 	return (ino_t)vino.ino;
579 }
580 
581 /* for printf-style formatting */
582 #define ceph_vinop(i) ceph_inode(i)->i_vino.ino, ceph_inode(i)->i_vino.snap
583 
ceph_ino(struct inode * inode)584 static inline u64 ceph_ino(struct inode *inode)
585 {
586 	return ceph_inode(inode)->i_vino.ino;
587 }
588 
ceph_snap(struct inode * inode)589 static inline u64 ceph_snap(struct inode *inode)
590 {
591 	return ceph_inode(inode)->i_vino.snap;
592 }
593 
594 /**
595  * ceph_present_ino - format an inode number for presentation to userland
596  * @sb: superblock where the inode lives
597  * @ino: inode number to (possibly) convert
598  *
599  * If the user mounted with the ino32 option, then the 64-bit value needs
600  * to be converted to something that can fit inside 32 bits. Note that
601  * internal kernel code never uses this value, so this is entirely for
602  * userland consumption.
603  */
ceph_present_ino(struct super_block * sb,u64 ino)604 static inline u64 ceph_present_ino(struct super_block *sb, u64 ino)
605 {
606 	if (unlikely(ceph_test_mount_opt(ceph_sb_to_fs_client(sb), INO32)))
607 		return ceph_ino_to_ino32(ino);
608 	return ino;
609 }
610 
ceph_present_inode(struct inode * inode)611 static inline u64 ceph_present_inode(struct inode *inode)
612 {
613 	return ceph_present_ino(inode->i_sb, ceph_ino(inode));
614 }
615 
ceph_ino_compare(struct inode * inode,void * data)616 static inline int ceph_ino_compare(struct inode *inode, void *data)
617 {
618 	struct ceph_vino *pvino = (struct ceph_vino *)data;
619 	struct ceph_inode_info *ci = ceph_inode(inode);
620 	return ci->i_vino.ino == pvino->ino &&
621 		ci->i_vino.snap == pvino->snap;
622 }
623 
624 /*
625  * The MDS reserves a set of inodes for its own usage. These should never
626  * be accessible by clients, and so the MDS has no reason to ever hand these
627  * out. The range is CEPH_MDS_INO_MDSDIR_OFFSET..CEPH_INO_SYSTEM_BASE.
628  *
629  * These come from src/mds/mdstypes.h in the ceph sources.
630  */
631 #define CEPH_MAX_MDS			0x100
632 #define CEPH_NUM_STRAY			10
633 #define CEPH_MDS_INO_MDSDIR_OFFSET	(1 * CEPH_MAX_MDS)
634 #define CEPH_MDS_INO_LOG_OFFSET		(2 * CEPH_MAX_MDS)
635 #define CEPH_INO_SYSTEM_BASE		((6*CEPH_MAX_MDS) + (CEPH_MAX_MDS * CEPH_NUM_STRAY))
636 
ceph_vino_is_reserved(const struct ceph_vino vino)637 static inline bool ceph_vino_is_reserved(const struct ceph_vino vino)
638 {
639 	if (vino.ino >= CEPH_INO_SYSTEM_BASE ||
640 	    vino.ino < CEPH_MDS_INO_MDSDIR_OFFSET)
641 		return false;
642 
643 	/* Don't warn on mdsdirs */
644 	WARN_RATELIMIT(vino.ino >= CEPH_MDS_INO_LOG_OFFSET,
645 			"Attempt to access reserved inode number 0x%llx",
646 			vino.ino);
647 	return true;
648 }
649 
ceph_find_inode(struct super_block * sb,struct ceph_vino vino)650 static inline struct inode *ceph_find_inode(struct super_block *sb,
651 					    struct ceph_vino vino)
652 {
653 	if (ceph_vino_is_reserved(vino))
654 		return NULL;
655 
656 	/*
657 	 * NB: The hashval will be run through the fs/inode.c hash function
658 	 * anyway, so there is no need to squash the inode number down to
659 	 * 32-bits first. Just use low-order bits on arches with 32-bit long.
660 	 */
661 	return ilookup5(sb, (unsigned long)vino.ino, ceph_ino_compare, &vino);
662 }
663 
664 
665 /*
666  * Ceph inode.
667  */
668 #define CEPH_I_DIR_ORDERED	(1 << 0)  /* dentries in dir are ordered */
669 #define CEPH_I_FLUSH		(1 << 2)  /* do not delay flush of dirty metadata */
670 #define CEPH_I_POOL_PERM	(1 << 3)  /* pool rd/wr bits are valid */
671 #define CEPH_I_POOL_RD		(1 << 4)  /* can read from pool */
672 #define CEPH_I_POOL_WR		(1 << 5)  /* can write to pool */
673 #define CEPH_I_SEC_INITED	(1 << 6)  /* security initialized */
674 #define CEPH_I_KICK_FLUSH	(1 << 7)  /* kick flushing caps */
675 #define CEPH_I_FLUSH_SNAPS	(1 << 8)  /* need flush snapss */
676 #define CEPH_I_ERROR_WRITE	(1 << 9) /* have seen write errors */
677 #define CEPH_I_ERROR_FILELOCK	(1 << 10) /* have seen file lock errors */
678 #define CEPH_I_ODIRECT_BIT	(11) /* inode in direct I/O mode */
679 #define CEPH_I_ODIRECT		(1 << CEPH_I_ODIRECT_BIT)
680 #define CEPH_ASYNC_CREATE_BIT	(12)	  /* async create in flight for this */
681 #define CEPH_I_ASYNC_CREATE	(1 << CEPH_ASYNC_CREATE_BIT)
682 #define CEPH_I_SHUTDOWN		(1 << 13) /* inode is no longer usable */
683 #define CEPH_I_ASYNC_CHECK_CAPS	(1 << 14) /* check caps immediately after async
684 					     creating finishes */
685 
686 /*
687  * Masks of ceph inode work.
688  */
689 #define CEPH_I_WORK_WRITEBACK		0
690 #define CEPH_I_WORK_INVALIDATE_PAGES	1
691 #define CEPH_I_WORK_VMTRUNCATE		2
692 #define CEPH_I_WORK_CHECK_CAPS		3
693 #define CEPH_I_WORK_FLUSH_SNAPS		4
694 
695 /*
696  * We set the ERROR_WRITE bit when we start seeing write errors on an inode
697  * and then clear it when they start succeeding. Note that we do a lockless
698  * check first, and only take the lock if it looks like it needs to be changed.
699  * The write submission code just takes this as a hint, so we're not too
700  * worried if a few slip through in either direction.
701  */
ceph_set_error_write(struct ceph_inode_info * ci)702 static inline void ceph_set_error_write(struct ceph_inode_info *ci)
703 {
704 	if (!(READ_ONCE(ci->i_ceph_flags) & CEPH_I_ERROR_WRITE)) {
705 		spin_lock(&ci->i_ceph_lock);
706 		ci->i_ceph_flags |= CEPH_I_ERROR_WRITE;
707 		spin_unlock(&ci->i_ceph_lock);
708 	}
709 }
710 
ceph_clear_error_write(struct ceph_inode_info * ci)711 static inline void ceph_clear_error_write(struct ceph_inode_info *ci)
712 {
713 	if (READ_ONCE(ci->i_ceph_flags) & CEPH_I_ERROR_WRITE) {
714 		spin_lock(&ci->i_ceph_lock);
715 		ci->i_ceph_flags &= ~CEPH_I_ERROR_WRITE;
716 		spin_unlock(&ci->i_ceph_lock);
717 	}
718 }
719 
__ceph_dir_set_complete(struct ceph_inode_info * ci,long long release_count,long long ordered_count)720 static inline void __ceph_dir_set_complete(struct ceph_inode_info *ci,
721 					   long long release_count,
722 					   long long ordered_count)
723 {
724 	/*
725 	 * Makes sure operations that setup readdir cache (update page
726 	 * cache and i_size) are strongly ordered w.r.t. the following
727 	 * atomic64_set() operations.
728 	 */
729 	smp_mb();
730 	atomic64_set(&ci->i_complete_seq[0], release_count);
731 	atomic64_set(&ci->i_complete_seq[1], ordered_count);
732 }
733 
__ceph_dir_clear_complete(struct ceph_inode_info * ci)734 static inline void __ceph_dir_clear_complete(struct ceph_inode_info *ci)
735 {
736 	atomic64_inc(&ci->i_release_count);
737 }
738 
__ceph_dir_clear_ordered(struct ceph_inode_info * ci)739 static inline void __ceph_dir_clear_ordered(struct ceph_inode_info *ci)
740 {
741 	atomic64_inc(&ci->i_ordered_count);
742 }
743 
__ceph_dir_is_complete(struct ceph_inode_info * ci)744 static inline bool __ceph_dir_is_complete(struct ceph_inode_info *ci)
745 {
746 	return atomic64_read(&ci->i_complete_seq[0]) ==
747 		atomic64_read(&ci->i_release_count);
748 }
749 
__ceph_dir_is_complete_ordered(struct ceph_inode_info * ci)750 static inline bool __ceph_dir_is_complete_ordered(struct ceph_inode_info *ci)
751 {
752 	return  atomic64_read(&ci->i_complete_seq[0]) ==
753 		atomic64_read(&ci->i_release_count) &&
754 		atomic64_read(&ci->i_complete_seq[1]) ==
755 		atomic64_read(&ci->i_ordered_count);
756 }
757 
ceph_dir_clear_complete(struct inode * inode)758 static inline void ceph_dir_clear_complete(struct inode *inode)
759 {
760 	__ceph_dir_clear_complete(ceph_inode(inode));
761 }
762 
ceph_dir_clear_ordered(struct inode * inode)763 static inline void ceph_dir_clear_ordered(struct inode *inode)
764 {
765 	__ceph_dir_clear_ordered(ceph_inode(inode));
766 }
767 
ceph_dir_is_complete_ordered(struct inode * inode)768 static inline bool ceph_dir_is_complete_ordered(struct inode *inode)
769 {
770 	bool ret = __ceph_dir_is_complete_ordered(ceph_inode(inode));
771 	smp_rmb();
772 	return ret;
773 }
774 
775 /* find a specific frag @f */
776 extern struct ceph_inode_frag *__ceph_find_frag(struct ceph_inode_info *ci,
777 						u32 f);
778 
779 /*
780  * choose fragment for value @v.  copy frag content to pfrag, if leaf
781  * exists
782  */
783 extern u32 ceph_choose_frag(struct ceph_inode_info *ci, u32 v,
784 			    struct ceph_inode_frag *pfrag,
785 			    int *found);
786 
ceph_dentry(const struct dentry * dentry)787 static inline struct ceph_dentry_info *ceph_dentry(const struct dentry *dentry)
788 {
789 	return (struct ceph_dentry_info *)dentry->d_fsdata;
790 }
791 
792 /*
793  * caps helpers
794  */
__ceph_is_any_real_caps(struct ceph_inode_info * ci)795 static inline bool __ceph_is_any_real_caps(struct ceph_inode_info *ci)
796 {
797 	return !RB_EMPTY_ROOT(&ci->i_caps);
798 }
799 
800 extern int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented);
801 extern int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int t);
802 extern int __ceph_caps_issued_mask_metric(struct ceph_inode_info *ci, int mask,
803 					  int t);
804 extern int __ceph_caps_issued_other(struct ceph_inode_info *ci,
805 				    struct ceph_cap *cap);
806 
ceph_caps_issued(struct ceph_inode_info * ci)807 static inline int ceph_caps_issued(struct ceph_inode_info *ci)
808 {
809 	int issued;
810 	spin_lock(&ci->i_ceph_lock);
811 	issued = __ceph_caps_issued(ci, NULL);
812 	spin_unlock(&ci->i_ceph_lock);
813 	return issued;
814 }
815 
ceph_caps_issued_mask_metric(struct ceph_inode_info * ci,int mask,int touch)816 static inline int ceph_caps_issued_mask_metric(struct ceph_inode_info *ci,
817 					       int mask, int touch)
818 {
819 	int r;
820 	spin_lock(&ci->i_ceph_lock);
821 	r = __ceph_caps_issued_mask_metric(ci, mask, touch);
822 	spin_unlock(&ci->i_ceph_lock);
823 	return r;
824 }
825 
__ceph_caps_dirty(struct ceph_inode_info * ci)826 static inline int __ceph_caps_dirty(struct ceph_inode_info *ci)
827 {
828 	return ci->i_dirty_caps | ci->i_flushing_caps;
829 }
830 extern struct ceph_cap_flush *ceph_alloc_cap_flush(void);
831 extern void ceph_free_cap_flush(struct ceph_cap_flush *cf);
832 extern int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask,
833 				  struct ceph_cap_flush **pcf);
834 
835 extern int __ceph_caps_revoking_other(struct ceph_inode_info *ci,
836 				      struct ceph_cap *ocap, int mask);
837 extern int __ceph_caps_used(struct ceph_inode_info *ci);
838 
__ceph_is_file_opened(struct ceph_inode_info * ci)839 static inline bool __ceph_is_file_opened(struct ceph_inode_info *ci)
840 {
841 	return ci->i_nr_by_mode[0];
842 }
843 extern int __ceph_caps_file_wanted(struct ceph_inode_info *ci);
844 extern int __ceph_caps_wanted(struct ceph_inode_info *ci);
845 
846 /* what the mds thinks we want */
847 extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci, bool check);
848 
849 extern void ceph_caps_init(struct ceph_mds_client *mdsc);
850 extern void ceph_caps_finalize(struct ceph_mds_client *mdsc);
851 extern void ceph_adjust_caps_max_min(struct ceph_mds_client *mdsc,
852 				     struct ceph_mount_options *fsopt);
853 extern int ceph_reserve_caps(struct ceph_mds_client *mdsc,
854 			     struct ceph_cap_reservation *ctx, int need);
855 extern void ceph_unreserve_caps(struct ceph_mds_client *mdsc,
856 			       struct ceph_cap_reservation *ctx);
857 extern void ceph_reservation_status(struct ceph_fs_client *client,
858 				    int *total, int *avail, int *used,
859 				    int *reserved, int *min);
860 extern void change_auth_cap_ses(struct ceph_inode_info *ci,
861 				struct ceph_mds_session *session);
862 
863 
864 
865 /*
866  * we keep buffered readdir results attached to file->private_data
867  */
868 #define CEPH_F_SYNC     1
869 #define CEPH_F_ATEND    2
870 
871 struct ceph_file_info {
872 	short fmode;     /* initialized on open */
873 	short flags;     /* CEPH_F_* */
874 
875 	spinlock_t rw_contexts_lock;
876 	struct list_head rw_contexts;
877 
878 	u32 filp_gen;
879 };
880 
881 struct ceph_dir_file_info {
882 	struct ceph_file_info file_info;
883 
884 	/* readdir: position within the dir */
885 	u32 frag;
886 	struct ceph_mds_request *last_readdir;
887 
888 	/* readdir: position within a frag */
889 	unsigned next_offset;  /* offset of next chunk (last_name's + 1) */
890 	char *last_name;       /* last entry in previous chunk */
891 	long long dir_release_count;
892 	long long dir_ordered_count;
893 	int readdir_cache_idx;
894 
895 	/* used for -o dirstat read() on directory thing */
896 	char *dir_info;
897 	int dir_info_len;
898 };
899 
900 struct ceph_rw_context {
901 	struct list_head list;
902 	struct task_struct *thread;
903 	int caps;
904 };
905 
906 #define CEPH_DEFINE_RW_CONTEXT(_name, _caps)	\
907 	struct ceph_rw_context _name = {	\
908 		.thread = current,		\
909 		.caps = _caps,			\
910 	}
911 
ceph_add_rw_context(struct ceph_file_info * cf,struct ceph_rw_context * ctx)912 static inline void ceph_add_rw_context(struct ceph_file_info *cf,
913 				       struct ceph_rw_context *ctx)
914 {
915 	spin_lock(&cf->rw_contexts_lock);
916 	list_add(&ctx->list, &cf->rw_contexts);
917 	spin_unlock(&cf->rw_contexts_lock);
918 }
919 
ceph_del_rw_context(struct ceph_file_info * cf,struct ceph_rw_context * ctx)920 static inline void ceph_del_rw_context(struct ceph_file_info *cf,
921 				       struct ceph_rw_context *ctx)
922 {
923 	spin_lock(&cf->rw_contexts_lock);
924 	list_del(&ctx->list);
925 	spin_unlock(&cf->rw_contexts_lock);
926 }
927 
928 static inline struct ceph_rw_context*
ceph_find_rw_context(struct ceph_file_info * cf)929 ceph_find_rw_context(struct ceph_file_info *cf)
930 {
931 	struct ceph_rw_context *ctx, *found = NULL;
932 	spin_lock(&cf->rw_contexts_lock);
933 	list_for_each_entry(ctx, &cf->rw_contexts, list) {
934 		if (ctx->thread == current) {
935 			found = ctx;
936 			break;
937 		}
938 	}
939 	spin_unlock(&cf->rw_contexts_lock);
940 	return found;
941 }
942 
943 struct ceph_readdir_cache_control {
944 	struct folio *folio;
945 	struct dentry **dentries;
946 	int index;
947 };
948 
949 /*
950  * A "snap realm" describes a subset of the file hierarchy sharing
951  * the same set of snapshots that apply to it.  The realms themselves
952  * are organized into a hierarchy, such that children inherit (some of)
953  * the snapshots of their parents.
954  *
955  * All inodes within the realm that have capabilities are linked into a
956  * per-realm list.
957  */
958 struct ceph_snap_realm {
959 	u64 ino;
960 	struct inode *inode;
961 	atomic_t nref;
962 	struct rb_node node;
963 
964 	u64 created, seq;
965 	u64 parent_ino;
966 	u64 parent_since;   /* snapid when our current parent became so */
967 
968 	u64 *prior_parent_snaps;      /* snaps inherited from any parents we */
969 	u32 num_prior_parent_snaps;   /*  had prior to parent_since */
970 	u64 *snaps;                   /* snaps specific to this realm */
971 	u32 num_snaps;
972 
973 	struct ceph_snap_realm *parent;
974 	struct list_head children;       /* list of child realms */
975 	struct list_head child_item;
976 
977 	struct list_head empty_item;     /* if i have ref==0 */
978 
979 	struct list_head dirty_item;     /* if realm needs new context */
980 
981 	struct list_head rebuild_item;   /* rebuild snap realms _downward_ in hierarchy */
982 
983 	/* the current set of snaps for this realm */
984 	struct ceph_snap_context *cached_context;
985 
986 	struct list_head inodes_with_caps;
987 	spinlock_t inodes_with_caps_lock;
988 };
989 
default_congestion_kb(void)990 static inline int default_congestion_kb(void)
991 {
992 	int congestion_kb;
993 
994 	/*
995 	 * Copied from NFS
996 	 *
997 	 * congestion size, scale with available memory.
998 	 *
999 	 *  64MB:    8192k
1000 	 * 128MB:   11585k
1001 	 * 256MB:   16384k
1002 	 * 512MB:   23170k
1003 	 *   1GB:   32768k
1004 	 *   2GB:   46340k
1005 	 *   4GB:   65536k
1006 	 *   8GB:   92681k
1007 	 *  16GB:  131072k
1008 	 *
1009 	 * This allows larger machines to have larger/more transfers.
1010 	 * Limit the default to 256M
1011 	 */
1012 	congestion_kb = (16*int_sqrt(totalram_pages())) << (PAGE_SHIFT-10);
1013 	if (congestion_kb > 256*1024)
1014 		congestion_kb = 256*1024;
1015 
1016 	return congestion_kb;
1017 }
1018 
1019 
1020 /* super.c */
1021 extern int ceph_force_reconnect(struct super_block *sb);
1022 /* snap.c */
1023 struct ceph_snap_realm *ceph_lookup_snap_realm(struct ceph_mds_client *mdsc,
1024 					       u64 ino);
1025 extern void ceph_get_snap_realm(struct ceph_mds_client *mdsc,
1026 				struct ceph_snap_realm *realm);
1027 extern void ceph_put_snap_realm(struct ceph_mds_client *mdsc,
1028 				struct ceph_snap_realm *realm);
1029 extern int ceph_update_snap_trace(struct ceph_mds_client *m,
1030 				  void *p, void *e, bool deletion,
1031 				  struct ceph_snap_realm **realm_ret);
1032 void ceph_change_snap_realm(struct inode *inode, struct ceph_snap_realm *realm);
1033 extern void ceph_handle_snap(struct ceph_mds_client *mdsc,
1034 			     struct ceph_mds_session *session,
1035 			     struct ceph_msg *msg);
1036 extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci,
1037 				  struct ceph_cap_snap *capsnap);
1038 extern void ceph_cleanup_global_and_empty_realms(struct ceph_mds_client *mdsc);
1039 
1040 extern struct ceph_snapid_map *ceph_get_snapid_map(struct ceph_mds_client *mdsc,
1041 						   u64 snap);
1042 extern void ceph_put_snapid_map(struct ceph_mds_client* mdsc,
1043 				struct ceph_snapid_map *sm);
1044 extern void ceph_trim_snapid_map(struct ceph_mds_client *mdsc);
1045 extern void ceph_cleanup_snapid_map(struct ceph_mds_client *mdsc);
1046 void ceph_umount_begin(struct super_block *sb);
1047 
1048 
1049 /*
1050  * a cap_snap is "pending" if it is still awaiting an in-progress
1051  * sync write (that may/may not still update size, mtime, etc.).
1052  */
__ceph_have_pending_cap_snap(struct ceph_inode_info * ci)1053 static inline bool __ceph_have_pending_cap_snap(struct ceph_inode_info *ci)
1054 {
1055 	return !list_empty(&ci->i_cap_snaps) &&
1056 	       list_last_entry(&ci->i_cap_snaps, struct ceph_cap_snap,
1057 			       ci_item)->writing;
1058 }
1059 
1060 /* inode.c */
1061 struct ceph_mds_reply_info_in;
1062 struct ceph_mds_reply_dirfrag;
1063 struct ceph_acl_sec_ctx;
1064 
1065 extern const struct inode_operations ceph_file_iops;
1066 
1067 extern struct inode *ceph_alloc_inode(struct super_block *sb);
1068 extern void ceph_evict_inode(struct inode *inode);
1069 extern void ceph_free_inode(struct inode *inode);
1070 
1071 struct inode *ceph_new_inode(struct inode *dir, struct dentry *dentry,
1072 			     umode_t *mode, struct ceph_acl_sec_ctx *as_ctx);
1073 void ceph_as_ctx_to_req(struct ceph_mds_request *req,
1074 			struct ceph_acl_sec_ctx *as_ctx);
1075 
1076 extern struct inode *ceph_get_inode(struct super_block *sb,
1077 				    struct ceph_vino vino,
1078 				    struct inode *newino);
1079 extern struct inode *ceph_get_snapdir(struct inode *parent);
1080 extern int ceph_fill_file_size(struct inode *inode, int issued,
1081 			       u32 truncate_seq, u64 truncate_size, u64 size);
1082 extern void ceph_inode_set_subvolume(struct inode *inode, u64 subvolume_id);
1083 extern void ceph_fill_file_time(struct inode *inode, int issued,
1084 				u64 time_warp_seq, struct timespec64 *ctime,
1085 				struct timespec64 *mtime,
1086 				struct timespec64 *atime);
1087 extern int ceph_fill_inode(struct inode *inode, struct page *locked_page,
1088 		    struct ceph_mds_reply_info_in *iinfo,
1089 		    struct ceph_mds_reply_dirfrag *dirinfo,
1090 		    struct ceph_mds_session *session, int cap_fmode,
1091 		    struct ceph_cap_reservation *caps_reservation);
1092 extern int ceph_fill_trace(struct super_block *sb,
1093 			   struct ceph_mds_request *req);
1094 extern int ceph_readdir_prepopulate(struct ceph_mds_request *req,
1095 				    struct ceph_mds_session *session);
1096 
1097 extern bool ceph_inode_set_size(struct inode *inode, loff_t size);
1098 extern void __ceph_do_pending_vmtruncate(struct inode *inode);
1099 
1100 void ceph_queue_inode_work(struct inode *inode, int work_bit);
1101 
ceph_queue_vmtruncate(struct inode * inode)1102 static inline void ceph_queue_vmtruncate(struct inode *inode)
1103 {
1104 	ceph_queue_inode_work(inode, CEPH_I_WORK_VMTRUNCATE);
1105 }
1106 
ceph_queue_invalidate(struct inode * inode)1107 static inline void ceph_queue_invalidate(struct inode *inode)
1108 {
1109 	ceph_queue_inode_work(inode, CEPH_I_WORK_INVALIDATE_PAGES);
1110 }
1111 
ceph_queue_writeback(struct inode * inode)1112 static inline void ceph_queue_writeback(struct inode *inode)
1113 {
1114 	ceph_queue_inode_work(inode, CEPH_I_WORK_WRITEBACK);
1115 }
1116 
ceph_queue_check_caps(struct inode * inode)1117 static inline void ceph_queue_check_caps(struct inode *inode)
1118 {
1119 	ceph_queue_inode_work(inode, CEPH_I_WORK_CHECK_CAPS);
1120 }
1121 
ceph_queue_flush_snaps(struct inode * inode)1122 static inline void ceph_queue_flush_snaps(struct inode *inode)
1123 {
1124 	ceph_queue_inode_work(inode, CEPH_I_WORK_FLUSH_SNAPS);
1125 }
1126 
1127 extern int ceph_try_to_choose_auth_mds(struct inode *inode, int mask);
1128 extern int __ceph_do_getattr(struct inode *inode, struct page *locked_page,
1129 			     int mask, bool force);
ceph_do_getattr(struct inode * inode,int mask,bool force)1130 static inline int ceph_do_getattr(struct inode *inode, int mask, bool force)
1131 {
1132 	return __ceph_do_getattr(inode, NULL, mask, force);
1133 }
1134 extern int ceph_permission(struct mnt_idmap *idmap,
1135 			   struct inode *inode, int mask);
1136 
1137 struct ceph_iattr {
1138 	struct ceph_fscrypt_auth	*fscrypt_auth;
1139 };
1140 
1141 extern int __ceph_setattr(struct mnt_idmap *idmap, struct inode *inode,
1142 			  struct iattr *attr, struct ceph_iattr *cia);
1143 extern int ceph_setattr(struct mnt_idmap *idmap,
1144 			struct dentry *dentry, struct iattr *attr);
1145 extern int ceph_getattr(struct mnt_idmap *idmap,
1146 			const struct path *path, struct kstat *stat,
1147 			u32 request_mask, unsigned int flags);
1148 void ceph_inode_shutdown(struct inode *inode);
1149 
ceph_inode_is_shutdown(struct inode * inode)1150 static inline bool ceph_inode_is_shutdown(struct inode *inode)
1151 {
1152 	unsigned long flags = READ_ONCE(ceph_inode(inode)->i_ceph_flags);
1153 	struct ceph_fs_client *fsc = ceph_inode_to_fs_client(inode);
1154 	int state = READ_ONCE(fsc->mount_state);
1155 
1156 	return (flags & CEPH_I_SHUTDOWN) || state >= CEPH_MOUNT_SHUTDOWN;
1157 }
1158 
1159 /* xattr.c */
1160 int __ceph_setxattr(struct inode *, const char *, const void *, size_t, int);
1161 int ceph_do_getvxattr(struct inode *inode, const char *name, void *value, size_t size);
1162 ssize_t __ceph_getxattr(struct inode *, const char *, void *, size_t);
1163 extern ssize_t ceph_listxattr(struct dentry *, char *, size_t);
1164 extern struct ceph_buffer *__ceph_build_xattrs_blob(struct ceph_inode_info *ci);
1165 extern void __ceph_destroy_xattrs(struct ceph_inode_info *ci);
1166 extern const struct xattr_handler * const ceph_xattr_handlers[];
1167 
1168 struct ceph_acl_sec_ctx {
1169 #ifdef CONFIG_CEPH_FS_POSIX_ACL
1170 	void *default_acl;
1171 	void *acl;
1172 #endif
1173 #ifdef CONFIG_CEPH_FS_SECURITY_LABEL
1174 	struct lsm_context lsmctx;
1175 #endif
1176 #ifdef CONFIG_FS_ENCRYPTION
1177 	struct ceph_fscrypt_auth *fscrypt_auth;
1178 #endif
1179 	struct ceph_pagelist *pagelist;
1180 };
1181 
1182 #ifdef CONFIG_SECURITY
1183 extern bool ceph_security_xattr_deadlock(struct inode *in);
1184 extern bool ceph_security_xattr_wanted(struct inode *in);
1185 #else
ceph_security_xattr_deadlock(struct inode * in)1186 static inline bool ceph_security_xattr_deadlock(struct inode *in)
1187 {
1188 	return false;
1189 }
ceph_security_xattr_wanted(struct inode * in)1190 static inline bool ceph_security_xattr_wanted(struct inode *in)
1191 {
1192 	return false;
1193 }
1194 #endif
1195 
1196 #ifdef CONFIG_CEPH_FS_SECURITY_LABEL
1197 extern int ceph_security_init_secctx(struct dentry *dentry, umode_t mode,
1198 				     struct ceph_acl_sec_ctx *ctx);
ceph_security_invalidate_secctx(struct inode * inode)1199 static inline void ceph_security_invalidate_secctx(struct inode *inode)
1200 {
1201 	security_inode_invalidate_secctx(inode);
1202 }
1203 #else
ceph_security_init_secctx(struct dentry * dentry,umode_t mode,struct ceph_acl_sec_ctx * ctx)1204 static inline int ceph_security_init_secctx(struct dentry *dentry, umode_t mode,
1205 					    struct ceph_acl_sec_ctx *ctx)
1206 {
1207 	return 0;
1208 }
ceph_security_invalidate_secctx(struct inode * inode)1209 static inline void ceph_security_invalidate_secctx(struct inode *inode)
1210 {
1211 }
1212 #endif
1213 
1214 void ceph_release_acl_sec_ctx(struct ceph_acl_sec_ctx *as_ctx);
1215 
1216 /* acl.c */
1217 #ifdef CONFIG_CEPH_FS_POSIX_ACL
1218 
1219 struct posix_acl *ceph_get_acl(struct inode *, int, bool);
1220 int ceph_set_acl(struct mnt_idmap *idmap,
1221 		 struct dentry *dentry, struct posix_acl *acl, int type);
1222 int ceph_pre_init_acls(struct inode *dir, umode_t *mode,
1223 		       struct ceph_acl_sec_ctx *as_ctx);
1224 void ceph_init_inode_acls(struct inode *inode,
1225 			  struct ceph_acl_sec_ctx *as_ctx);
1226 
ceph_forget_all_cached_acls(struct inode * inode)1227 static inline void ceph_forget_all_cached_acls(struct inode *inode)
1228 {
1229        forget_all_cached_acls(inode);
1230 }
1231 
1232 #else
1233 
1234 #define ceph_get_acl NULL
1235 #define ceph_set_acl NULL
1236 
ceph_pre_init_acls(struct inode * dir,umode_t * mode,struct ceph_acl_sec_ctx * as_ctx)1237 static inline int ceph_pre_init_acls(struct inode *dir, umode_t *mode,
1238 				     struct ceph_acl_sec_ctx *as_ctx)
1239 {
1240 	return 0;
1241 }
ceph_init_inode_acls(struct inode * inode,struct ceph_acl_sec_ctx * as_ctx)1242 static inline void ceph_init_inode_acls(struct inode *inode,
1243 					struct ceph_acl_sec_ctx *as_ctx)
1244 {
1245 }
1246 
ceph_forget_all_cached_acls(struct inode * inode)1247 static inline void ceph_forget_all_cached_acls(struct inode *inode)
1248 {
1249 }
1250 
1251 #endif
1252 
1253 /* caps.c */
1254 extern const char *ceph_cap_string(int c);
1255 extern void ceph_handle_caps(struct ceph_mds_session *session,
1256 			     struct ceph_msg *msg);
1257 extern struct ceph_cap *ceph_get_cap(struct ceph_mds_client *mdsc,
1258 				     struct ceph_cap_reservation *ctx);
1259 extern void ceph_add_cap(struct inode *inode,
1260 			 struct ceph_mds_session *session, u64 cap_id,
1261 			 unsigned issued, unsigned wanted,
1262 			 unsigned cap, unsigned seq, u64 realmino, int flags,
1263 			 struct ceph_cap **new_cap);
1264 extern void __ceph_remove_cap(struct ceph_cap *cap, bool queue_release);
1265 extern void ceph_remove_cap(struct ceph_mds_client *mdsc, struct ceph_cap *cap,
1266 			    bool queue_release);
1267 extern void __ceph_remove_caps(struct ceph_inode_info *ci);
1268 extern void ceph_put_cap(struct ceph_mds_client *mdsc,
1269 			 struct ceph_cap *cap);
1270 extern int ceph_is_any_caps(struct inode *inode);
1271 
1272 extern int ceph_write_inode(struct inode *inode, struct writeback_control *wbc);
1273 extern int ceph_fsync(struct file *file, loff_t start, loff_t end,
1274 		      int datasync);
1275 extern void ceph_early_kick_flushing_caps(struct ceph_mds_client *mdsc,
1276 					  struct ceph_mds_session *session);
1277 extern void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc,
1278 				    struct ceph_mds_session *session);
1279 void ceph_kick_flushing_inode_caps(struct ceph_mds_session *session,
1280 				   struct ceph_inode_info *ci);
1281 extern struct ceph_cap *__get_cap_for_mds(struct ceph_inode_info *ci,
1282 					  int mds);
1283 extern struct ceph_cap *ceph_get_cap_for_mds(struct ceph_inode_info *ci,
1284 					     int mds);
1285 extern void ceph_take_cap_refs(struct ceph_inode_info *ci, int caps,
1286 				bool snap_rwsem_locked);
1287 extern void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps);
1288 extern void ceph_put_cap_refs(struct ceph_inode_info *ci, int had);
1289 extern void ceph_put_cap_refs_async(struct ceph_inode_info *ci, int had);
1290 extern void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr,
1291 				       struct ceph_snap_context *snapc);
1292 extern void __ceph_remove_capsnap(struct inode *inode,
1293 				  struct ceph_cap_snap *capsnap,
1294 				  bool *wake_ci, bool *wake_mdsc);
1295 extern void ceph_remove_capsnap(struct inode *inode,
1296 				struct ceph_cap_snap *capsnap,
1297 				bool *wake_ci, bool *wake_mdsc);
1298 extern void ceph_flush_snaps(struct ceph_inode_info *ci,
1299 			     struct ceph_mds_session **psession);
1300 extern bool __ceph_should_report_size(struct ceph_inode_info *ci);
1301 extern void ceph_check_caps(struct ceph_inode_info *ci, int flags);
1302 extern unsigned long ceph_check_delayed_caps(struct ceph_mds_client *mdsc);
1303 extern void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc);
1304 extern void ceph_flush_cap_releases(struct ceph_mds_client *mdsc);
1305 extern int  ceph_drop_caps_for_unlink(struct inode *inode);
1306 extern int ceph_encode_inode_release(void **p, struct inode *inode,
1307 				     int mds, int drop, int unless, int force);
1308 extern int ceph_encode_dentry_release(void **p, struct dentry *dn,
1309 				      struct inode *dir,
1310 				      int mds, int drop, int unless);
1311 
1312 extern int __ceph_get_caps(struct inode *inode, struct ceph_file_info *fi,
1313 			   int need, int want, loff_t endoff, int *got);
1314 extern int ceph_get_caps(struct file *filp, int need, int want,
1315 			 loff_t endoff, int *got);
1316 extern int ceph_try_get_caps(struct inode *inode,
1317 			     int need, int want, bool nonblock, int *got);
1318 
1319 /* for counting open files by mode */
1320 extern void ceph_get_fmode(struct ceph_inode_info *ci, int mode, int count);
1321 extern void ceph_put_fmode(struct ceph_inode_info *ci, int mode, int count);
1322 extern void __ceph_touch_fmode(struct ceph_inode_info *ci,
1323 			       struct ceph_mds_client *mdsc, int fmode);
1324 
1325 /* addr.c */
1326 extern const struct address_space_operations ceph_aops;
1327 extern const struct netfs_request_ops ceph_netfs_ops;
1328 int ceph_mmap_prepare(struct vm_area_desc *desc);
1329 extern int ceph_uninline_data(struct file *file);
1330 extern int ceph_pool_perm_check(struct inode *inode, int need);
1331 extern void ceph_pool_perm_destroy(struct ceph_mds_client* mdsc);
1332 int ceph_purge_inode_cap(struct inode *inode, struct ceph_cap *cap, bool *invalidate);
1333 
ceph_has_inline_data(struct ceph_inode_info * ci)1334 static inline bool ceph_has_inline_data(struct ceph_inode_info *ci)
1335 {
1336 	if (ci->i_inline_version == CEPH_INLINE_NONE ||
1337 	    ci->i_inline_version == 1) /* initial version, no data */
1338 		return false;
1339 	return true;
1340 }
1341 
1342 /* file.c */
1343 extern const struct file_operations ceph_file_fops;
1344 
1345 extern int ceph_renew_caps(struct inode *inode, int fmode);
1346 extern int ceph_open(struct inode *inode, struct file *file);
1347 extern int ceph_atomic_open(struct inode *dir, struct dentry *dentry,
1348 			    struct file *file, unsigned flags, umode_t mode);
1349 extern ssize_t __ceph_sync_read(struct inode *inode, loff_t *ki_pos,
1350 				struct iov_iter *to, int *retry_op,
1351 				u64 *last_objver);
1352 extern int ceph_release(struct inode *inode, struct file *filp);
1353 extern void ceph_fill_inline_data(struct inode *inode, struct page *locked_page,
1354 				  char *data, size_t len);
1355 
1356 /* dir.c */
1357 extern const struct file_operations ceph_dir_fops;
1358 extern const struct file_operations ceph_snapdir_fops;
1359 extern const struct inode_operations ceph_dir_iops;
1360 extern const struct inode_operations ceph_snapdir_iops;
1361 extern const struct dentry_operations ceph_dentry_ops;
1362 
1363 extern loff_t ceph_make_fpos(unsigned high, unsigned off, bool hash_order);
1364 extern int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry);
1365 extern struct dentry *ceph_handle_snapdir(struct ceph_mds_request *req,
1366 			       struct dentry *dentry);
1367 extern struct dentry *ceph_finish_lookup(struct ceph_mds_request *req,
1368 					 struct dentry *dentry, int err);
1369 
1370 extern void __ceph_dentry_lease_touch(struct ceph_dentry_info *di);
1371 extern void __ceph_dentry_dir_lease_touch(struct ceph_dentry_info *di);
1372 extern void ceph_invalidate_dentry_lease(struct dentry *dentry);
1373 extern int ceph_trim_dentries(struct ceph_mds_client *mdsc);
1374 extern unsigned ceph_dentry_hash(struct inode *dir, struct dentry *dn);
1375 extern void ceph_readdir_cache_release(struct ceph_readdir_cache_control *ctl);
1376 
1377 /* ioctl.c */
1378 extern long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
1379 
1380 /* export.c */
1381 extern const struct export_operations ceph_export_ops;
1382 struct inode *ceph_lookup_inode(struct super_block *sb, u64 ino);
1383 
1384 /* locks.c */
1385 extern __init void ceph_flock_init(void);
1386 extern int ceph_lock(struct file *file, int cmd, struct file_lock *fl);
1387 extern int ceph_flock(struct file *file, int cmd, struct file_lock *fl);
1388 extern void ceph_count_locks(struct inode *inode, int *p_num, int *f_num);
1389 extern int ceph_encode_locks_to_buffer(struct inode *inode,
1390 				       struct ceph_filelock *flocks,
1391 				       int num_fcntl_locks,
1392 				       int num_flock_locks);
1393 extern int ceph_locks_to_pagelist(struct ceph_filelock *flocks,
1394 				  struct ceph_pagelist *pagelist,
1395 				  int num_fcntl_locks, int num_flock_locks);
1396 
1397 /* debugfs.c */
1398 extern void ceph_fs_debugfs_init(struct ceph_fs_client *client);
1399 extern void ceph_fs_debugfs_cleanup(struct ceph_fs_client *client);
1400 
1401 /* quota.c */
1402 
1403 enum quota_get_realm {
1404 	QUOTA_GET_MAX_FILES,
1405 	QUOTA_GET_MAX_BYTES,
1406 	QUOTA_GET_ANY
1407 };
1408 
__ceph_has_quota(struct ceph_inode_info * ci,enum quota_get_realm which)1409 static inline bool __ceph_has_quota(struct ceph_inode_info *ci,
1410 				    enum quota_get_realm which)
1411 {
1412 	bool has_quota = false;
1413 
1414 	switch (which) {
1415 	case QUOTA_GET_MAX_BYTES:
1416 		has_quota = !!ci->i_max_bytes;
1417 		break;
1418 	case QUOTA_GET_MAX_FILES:
1419 		has_quota = !!ci->i_max_files;
1420 		break;
1421 	default:
1422 		has_quota = !!(ci->i_max_files || ci->i_max_bytes);
1423 	}
1424 	return has_quota;
1425 }
1426 
1427 extern void ceph_adjust_quota_realms_count(struct inode *inode, bool inc);
1428 
__ceph_update_quota(struct ceph_inode_info * ci,u64 max_bytes,u64 max_files)1429 static inline void __ceph_update_quota(struct ceph_inode_info *ci,
1430 				       u64 max_bytes, u64 max_files)
1431 {
1432 	bool had_quota, has_quota;
1433 	had_quota = __ceph_has_quota(ci, QUOTA_GET_ANY);
1434 	ci->i_max_bytes = max_bytes;
1435 	ci->i_max_files = max_files;
1436 	has_quota = __ceph_has_quota(ci, QUOTA_GET_ANY);
1437 
1438 	if (had_quota != has_quota)
1439 		ceph_adjust_quota_realms_count(&ci->netfs.inode, has_quota);
1440 }
1441 
__ceph_sparse_read_ext_count(struct inode * inode,u64 len)1442 static inline int __ceph_sparse_read_ext_count(struct inode *inode, u64 len)
1443 {
1444 	int cnt = 0;
1445 
1446 	if (IS_ENCRYPTED(inode)) {
1447 		cnt = len >> CEPH_FSCRYPT_BLOCK_SHIFT;
1448 		if (cnt > CEPH_SPARSE_EXT_ARRAY_INITIAL)
1449 			cnt = 0;
1450 	}
1451 
1452 	return cnt;
1453 }
1454 
1455 extern void ceph_handle_quota(struct ceph_mds_client *mdsc,
1456 			      struct ceph_mds_session *session,
1457 			      struct ceph_msg *msg);
1458 extern bool ceph_quota_is_max_files_exceeded(struct inode *inode);
1459 extern bool ceph_quota_is_same_realm(struct inode *old, struct inode *new);
1460 extern bool ceph_quota_is_max_bytes_exceeded(struct inode *inode,
1461 					     loff_t newlen);
1462 extern bool ceph_quota_is_max_bytes_approaching(struct inode *inode,
1463 						loff_t newlen);
1464 extern bool ceph_quota_update_statfs(struct ceph_fs_client *fsc,
1465 				     struct kstatfs *buf);
1466 extern void ceph_cleanup_quotarealms_inodes(struct ceph_mds_client *mdsc);
1467 
1468 bool ceph_inc_mds_stopping_blocker(struct ceph_mds_client *mdsc,
1469 			       struct ceph_mds_session *session);
1470 void ceph_dec_mds_stopping_blocker(struct ceph_mds_client *mdsc);
1471 bool ceph_inc_osd_stopping_blocker(struct ceph_mds_client *mdsc);
1472 void ceph_dec_osd_stopping_blocker(struct ceph_mds_client *mdsc);
1473 #endif /* _FS_CEPH_SUPER_H */
1474