xref: /openbmc/linux/fs/ceph/mds_client.h (revision 4f2c0a4acffbec01079c28f839422e64ddeff004)
1b2441318SGreg Kroah-Hartman /* SPDX-License-Identifier: GPL-2.0 */
22f2dc053SSage Weil #ifndef _FS_CEPH_MDS_CLIENT_H
32f2dc053SSage Weil #define _FS_CEPH_MDS_CLIENT_H
42f2dc053SSage Weil 
52f2dc053SSage Weil #include <linux/completion.h>
6153c8e6bSSage Weil #include <linux/kref.h>
72f2dc053SSage Weil #include <linux/list.h>
82f2dc053SSage Weil #include <linux/mutex.h>
944ca18f2SSage Weil #include <linux/rbtree.h>
102f2dc053SSage Weil #include <linux/spinlock.h>
113997c01dSElena Reshetova #include <linux/refcount.h>
12717e6f28SYan, Zheng #include <linux/utsname.h>
1370c94820SXiubo Li #include <linux/ktime.h>
142f2dc053SSage Weil 
153d14c5d2SYehuda Sadeh #include <linux/ceph/types.h>
163d14c5d2SYehuda Sadeh #include <linux/ceph/messenger.h>
173d14c5d2SYehuda Sadeh #include <linux/ceph/mdsmap.h>
186c4a1915SAlex Elder #include <linux/ceph/auth.h>
192f2dc053SSage Weil 
20f9009efaSXiubo Li #include "metric.h"
213e699bd8SXiubo Li #include "super.h"
22f9009efaSXiubo Li 
23342ce182SYan, Zheng /* The first 8 bits are reserved for old ceph releases */
249ba1e224SXiubo Li enum ceph_feature_type {
259ba1e224SXiubo Li 	CEPHFS_FEATURE_MIMIC = 8,
269ba1e224SXiubo Li 	CEPHFS_FEATURE_REPLY_ENCODING,
279ba1e224SXiubo Li 	CEPHFS_FEATURE_RECLAIM_CLIENT,
289ba1e224SXiubo Li 	CEPHFS_FEATURE_LAZY_CAP_WANTED,
299ba1e224SXiubo Li 	CEPHFS_FEATURE_MULTI_RECONNECT,
30d4846487SJeff Layton 	CEPHFS_FEATURE_DELEG_INO,
3118f473b3SXiubo Li 	CEPHFS_FEATURE_METRIC_COLLECT,
32300e42a2SXiubo Li 	CEPHFS_FEATURE_ALTERNATE_NAME,
33300e42a2SXiubo Li 	CEPHFS_FEATURE_NOTIFY_SESSION_STATE,
34*6eb06c46SXiubo Li 	CEPHFS_FEATURE_OP_GETVXATTR,
35342ce182SYan, Zheng 	CEPHFS_FEATURE_32BITS_RETRY_FWD,
36*6eb06c46SXiubo Li 
379ba1e224SXiubo Li 	CEPHFS_FEATURE_MAX = CEPHFS_FEATURE_32BITS_RETRY_FWD,
389ba1e224SXiubo Li };
39d2f8bb27SYan, Zheng 
40342ce182SYan, Zheng #define CEPHFS_FEATURES_CLIENT_SUPPORTED {	\
41342ce182SYan, Zheng 	0, 1, 2, 3, 4, 5, 6, 7,			\
42b37fe1f9SYan, Zheng 	CEPHFS_FEATURE_MIMIC,			\
43d2f8bb27SYan, Zheng 	CEPHFS_FEATURE_REPLY_ENCODING,		\
4481c5a148SYan, Zheng 	CEPHFS_FEATURE_LAZY_CAP_WANTED,		\
45d4846487SJeff Layton 	CEPHFS_FEATURE_MULTI_RECONNECT,		\
4618f473b3SXiubo Li 	CEPHFS_FEATURE_DELEG_INO,		\
47300e42a2SXiubo Li 	CEPHFS_FEATURE_METRIC_COLLECT,		\
48*6eb06c46SXiubo Li 	CEPHFS_FEATURE_ALTERNATE_NAME,		\
49342ce182SYan, Zheng 	CEPHFS_FEATURE_NOTIFY_SESSION_STATE,	\
50342ce182SYan, Zheng 	CEPHFS_FEATURE_OP_GETVXATTR,		\
512f2dc053SSage Weil 	CEPHFS_FEATURE_32BITS_RETRY_FWD,	\
522f2dc053SSage Weil }
532f2dc053SSage Weil 
542f2dc053SSage Weil /*
552f2dc053SSage Weil  * Some lock dependencies:
562f2dc053SSage Weil  *
572f2dc053SSage Weil  * session->s_mutex
582f2dc053SSage Weil  *         mdsc->mutex
59be655596SSage Weil  *
602f2dc053SSage Weil  *         mdsc->snap_rwsem
612f2dc053SSage Weil  *
622f2dc053SSage Weil  *         ci->i_ceph_lock
632f2dc053SSage Weil  *                 mdsc->snap_flush_lock
642f2dc053SSage Weil  *                 mdsc->cap_delay_lock
653d14c5d2SYehuda Sadeh  *
662f2dc053SSage Weil  */
672f2dc053SSage Weil 
682f2dc053SSage Weil struct ceph_fs_client;
692f2dc053SSage Weil struct ceph_cap;
702f2dc053SSage Weil 
712f2dc053SSage Weil /*
722f2dc053SSage Weil  * parsed info about a single inode.  pointers are into the encoded
732f2dc053SSage Weil  * on-wire structures within the mds reply message payload.
7414303d20SSage Weil  */
752f2dc053SSage Weil struct ceph_mds_reply_info_in {
762f2dc053SSage Weil 	struct ceph_mds_reply_inode *in;
772f2dc053SSage Weil 	struct ceph_dir_layout dir_layout;
782f2dc053SSage Weil 	u32 symlink_len;
79fb01d1f8SYan, Zheng 	char *symlink;
80fb01d1f8SYan, Zheng 	u32 xattr_len;
81fb01d1f8SYan, Zheng 	char *xattr_data;
825ea5c5e0SYan, Zheng 	u64 inline_version;
83779fe0fbSYan, Zheng 	u32 inline_len;
84fb18a575SLuis Henriques 	char *inline_data;
85fb18a575SLuis Henriques 	u32 pool_ns_len;
8608796873SYan, Zheng 	char *pool_ns_data;
87245ce991SJeff Layton 	u64 max_bytes;
88193e7b37SDavid Disseldorp 	u64 max_files;
89e7f72952SYanhu Cao 	s32 dir_pin;
90a35ead31SJeff Layton 	struct ceph_timespec btime;
912f2dc053SSage Weil 	struct ceph_timespec snap_btime;
922f2dc053SSage Weil 	u8 *fscrypt_auth;
932a5beea3SYan, Zheng 	u8 *fscrypt_file;
942a5beea3SYan, Zheng 	u32 fscrypt_auth_len;
952a5beea3SYan, Zheng 	u32 fscrypt_file_len;
962a5beea3SYan, Zheng 	u64 rsnaps;
972a5beea3SYan, Zheng 	u64 change_attr;
988974eebdSYan, Zheng };
992a5beea3SYan, Zheng 
1002a5beea3SYan, Zheng struct ceph_mds_reply_dir_entry {
1016ddf5f16SMilind Changire 	bool			      is_nokey;
1026ddf5f16SMilind Changire 	char                          *name;
1036ddf5f16SMilind Changire 	u32                           name_len;
1046ddf5f16SMilind Changire 	u32			      raw_hash;
1056ddf5f16SMilind Changire 	struct ceph_mds_reply_lease   *lease;
1062f2dc053SSage Weil 	struct ceph_mds_reply_info_in inode;
10725933abdSHerb Shiu 	loff_t			      offset;
10825933abdSHerb Shiu };
10925933abdSHerb Shiu 
11025933abdSHerb Shiu struct ceph_mds_reply_xattr {
1112f2dc053SSage Weil 	char *xattr_value;
1122f2dc053SSage Weil 	size_t xattr_value_len;
1132f2dc053SSage Weil };
1142f2dc053SSage Weil 
11525933abdSHerb Shiu /*
1162f2dc053SSage Weil  * parsed info about an mds reply, including information about
1172f2dc053SSage Weil  * either: 1) the target inode and/or its parent directory and dentry,
1182f2dc053SSage Weil  * and directory contents (for readdir results), or
1192f2dc053SSage Weil  * 2) the file range lock info (for fcntl F_GETLK results).
1202f2dc053SSage Weil  */
1216ddf5f16SMilind Changire struct ceph_mds_reply_info_parsed {
1222f2dc053SSage Weil 	struct ceph_mds_reply_head    *head;
12325933abdSHerb Shiu 
12425933abdSHerb Shiu 	/* trace */
12525933abdSHerb Shiu 	struct ceph_mds_reply_info_in diri, targeti;
12625933abdSHerb Shiu 	struct ceph_mds_reply_dirfrag *dirfrag;
12725933abdSHerb Shiu 	char                          *dname;
12825933abdSHerb Shiu 	u8			      *altname;
12925933abdSHerb Shiu 	u32                           dname_len;
1302f2dc053SSage Weil 	u32                           altname_len;
13154008399SYan, Zheng 	struct ceph_mds_reply_lease   *dlease;
1322f2dc053SSage Weil 	struct ceph_mds_reply_xattr   xattr_info;
133f3c4ebe6SYan, Zheng 
13479162547SYan, Zheng 	/* extra */
135f3c4ebe6SYan, Zheng 	union {
13679162547SYan, Zheng 		/* for fcntl F_GETLK results */
1372a5beea3SYan, Zheng 		struct ceph_filelock *filelock_reply;
13825933abdSHerb Shiu 
1396e8575faSSam Lang 		/* for readdir results */
1406e8575faSSam Lang 		struct {
1416e8575faSSam Lang 			struct ceph_mds_reply_dirfrag *dir_dir;
1426e8575faSSam Lang 			size_t			      dir_buf_size;
1436e8575faSSam Lang 			int                           dir_nr;
1446e8575faSSam Lang 			bool			      dir_end;
14525933abdSHerb Shiu 			bool			      dir_complete;
1462f2dc053SSage Weil 			bool			      hash_order;
1472f2dc053SSage Weil 			bool			      offset_hash;
1482f2dc053SSage Weil 			struct ceph_mds_reply_dir_entry  *dir_entries;
1492f2dc053SSage Weil 		};
1502f2dc053SSage Weil 
1512f2dc053SSage Weil 		/* for create results */
1522f2dc053SSage Weil 		struct {
1532f2dc053SSage Weil 			bool has_create_ino;
1542f2dc053SSage Weil 			u64 ino;
1552f2dc053SSage Weil 		};
15692475f05SJeff Layton 	};
15792475f05SJeff Layton 
15892475f05SJeff Layton 	/* encoded blob describing snapshot contexts for certain
1592f2dc053SSage Weil 	   operations (e.g., open) */
16092475f05SJeff Layton 	void *snapblob;
1612f2dc053SSage Weil 	int snapblob_len;
1622f2dc053SSage Weil };
1632f2dc053SSage Weil 
1642f2dc053SSage Weil 
1652f2dc053SSage Weil /*
1662f2dc053SSage Weil  * cap releases are batched and sent to the MDS en masse.
1672f2dc053SSage Weil  *
1682f2dc053SSage Weil  * Account for per-message overhead of mds_cap_release header
1692f2dc053SSage Weil  * and __le32 for osd epoch barrier trailing field.
1702f2dc053SSage Weil  */
1712f2dc053SSage Weil #define CEPH_CAPS_PER_RELEASE ((PAGE_SIZE - sizeof(u32) -		\
1722f2dc053SSage Weil 				sizeof(struct ceph_mds_cap_release)) /	\
1737e6906c1SYan, Zheng 			        sizeof(struct ceph_mds_cap_item))
1747e6906c1SYan, Zheng 
1757e6906c1SYan, Zheng 
1764d681c2fSXiubo Li /*
1774d681c2fSXiubo Li  * state associated with each MDS<->client session
1782f2dc053SSage Weil  */
1792f2dc053SSage Weil enum {
1802f2dc053SSage Weil 	CEPH_MDS_SESSION_NEW = 1,
1812f2dc053SSage Weil 	CEPH_MDS_SESSION_OPENING = 2,
1822f2dc053SSage Weil 	CEPH_MDS_SESSION_OPEN = 3,
1832f2dc053SSage Weil 	CEPH_MDS_SESSION_HUNG = 4,
1842f2dc053SSage Weil 	CEPH_MDS_SESSION_RESTARTING = 5,
18584bf3950SYan, Zheng 	CEPH_MDS_SESSION_RECONNECTING = 6,
1862f2dc053SSage Weil 	CEPH_MDS_SESSION_CLOSING = 7,
1872f2dc053SSage Weil 	CEPH_MDS_SESSION_CLOSED = 8,
1882f2dc053SSage Weil 	CEPH_MDS_SESSION_REJECTED = 9,
1892f2dc053SSage Weil };
1902f2dc053SSage Weil 
1916c4a1915SAlex Elder struct ceph_mds_session {
1924e7a5dcdSSage Weil 	struct ceph_mds_client *s_mdsc;
19352d60f8eSJeff Layton 	int               s_mds;
19452d60f8eSJeff Layton 	int               s_state;
195d8fb02abSAlex Elder 	unsigned long     s_ttl;      /* time until mds kills us */
196d8fb02abSAlex Elder 	unsigned long	  s_features;
197d8fb02abSAlex Elder 	u64               s_seq;      /* incoming msg seq # */
198045100cdSJeff Layton 	struct mutex      s_mutex;    /* serialize session messages */
1992f2dc053SSage Weil 
200e3ec8d68SYan, Zheng 	struct ceph_connection s_con;
201533a2818SJeff Layton 
2022f2dc053SSage Weil 	struct ceph_auth_handshake s_auth;
20399a9c273SYan, Zheng 
20403f4fcb0SYan, Zheng 	atomic_t          s_cap_gen;  /* inc each time we get mds stale msg */
2052f2dc053SSage Weil 	unsigned long     s_cap_ttl;  /* when session caps expire. protected by s_mutex */
206e3ec8d68SYan, Zheng 
2072f2dc053SSage Weil 	/* protected by s_cap_lock */
208829ad4dbSJeff Layton 	spinlock_t        s_cap_lock;
2091cf03a68SJeff Layton 	refcount_t        s_ref;
210829ad4dbSJeff Layton 	struct list_head  s_caps;     /* all caps issued by this session */
211829ad4dbSJeff Layton 	struct ceph_cap  *s_cap_iterator;
2122f2dc053SSage Weil 	int               s_nr_caps;
2131cf03a68SJeff Layton 	int               s_num_cap_releases;
2142f2dc053SSage Weil 	int		  s_cap_reconnect;
2152f2dc053SSage Weil 	int		  s_readonly;
2162f2dc053SSage Weil 	struct list_head  s_cap_releases; /* waiting cap_release messages */
2172f2dc053SSage Weil 	struct work_struct s_cap_release_work;
2182f2dc053SSage Weil 
219d4846487SJeff Layton 	/* See ceph_inode_info->i_dirty_item. */
2202f2dc053SSage Weil 	struct list_head  s_cap_dirty;	      /* inodes w/ dirty caps */
2212f2dc053SSage Weil 
2222f2dc053SSage Weil 	/* See ceph_inode_info->i_flushing_item. */
2232f2dc053SSage Weil 	struct list_head  s_cap_flushing;     /* inodes w/ flushing caps */
2242f2dc053SSage Weil 
2252f2dc053SSage Weil 	unsigned long     s_renew_requested; /* last time we sent a renew req */
2262f2dc053SSage Weil 	u64               s_renew_seq;
2272f2dc053SSage Weil 
2282f2dc053SSage Weil 	struct list_head  s_waiting;  /* waiting requests */
2292f2dc053SSage Weil 	struct list_head  s_unsafe;   /* unsafe requests */
2302f2dc053SSage Weil 	struct xarray	  s_delegated_inos;
2312f2dc053SSage Weil };
2322f2dc053SSage Weil 
2332f2dc053SSage Weil /*
2342f2dc053SSage Weil  * modes of choosing which MDS to send a request to
2352f2dc053SSage Weil  */
2362f2dc053SSage Weil enum {
2372f2dc053SSage Weil 	USE_ANY_MDS,
2382f2dc053SSage Weil 	USE_RANDOM_MDS,
2399280be24SYan, Zheng 	USE_AUTH_MDS,   /* prefer authoritative mds for this metadata item */
2409280be24SYan, Zheng };
2419280be24SYan, Zheng 
2429280be24SYan, Zheng struct ceph_mds_request;
2439280be24SYan, Zheng struct ceph_mds_client;
2442f2dc053SSage Weil 
2452f2dc053SSage Weil /*
2462f2dc053SSage Weil  * request completion callback
2472f2dc053SSage Weil  */
2482f2dc053SSage Weil typedef void (*ceph_mds_request_callback_t) (struct ceph_mds_client *mdsc,
2492f2dc053SSage Weil 					     struct ceph_mds_request *req);
25044ca18f2SSage Weil /*
25137151668SYehuda Sadeh  * wait for request completion callback
2522f2dc053SSage Weil  */
253045100cdSJeff Layton typedef int (*ceph_mds_request_wait_callback_t) (struct ceph_mds_client *mdsc,
2542f2dc053SSage Weil 						 struct ceph_mds_request *req);
2552f2dc053SSage Weil 
2562f2dc053SSage Weil /*
2572f2dc053SSage Weil  * an in-flight mds request
2582f2dc053SSage Weil  */
2592f2dc053SSage Weil struct ceph_mds_request {
26041b02e1fSSage Weil 	u64 r_tid;                   /* transaction id */
2612f2dc053SSage Weil 	struct rb_node r_node;
2622f2dc053SSage Weil 	struct ceph_mds_client *r_mdsc;
2632f2dc053SSage Weil 
2643dd69aabSJeff Layton 	struct kref       r_kref;
2652f2dc053SSage Weil 	int r_op;                    /* mds op code */
2662f2dc053SSage Weil 
267bc2de10dSJeff Layton 	/* operation on what? */
268bc2de10dSJeff Layton 	struct inode *r_inode;              /* arg1 */
269bc2de10dSJeff Layton 	struct dentry *r_dentry;            /* arg1 */
270bc2de10dSJeff Layton 	struct dentry *r_old_dentry;        /* arg2: rename from or link from */
271bc2de10dSJeff Layton 	struct inode *r_old_dentry_dir;     /* arg2: old dentry's parent dir */
272bc2de10dSJeff Layton 	char *r_path1, *r_path2;
2733dd69aabSJeff Layton 	struct ceph_vino r_ino1, r_ino2;
2743bb48b41SJeff Layton 
275bc2de10dSJeff Layton 	struct inode *r_parent;		    /* parent dir inode */
276bc2de10dSJeff Layton 	struct inode *r_target_inode;       /* resulting inode */
277b4556396SSage Weil 	struct inode *r_new_inode;	    /* new inode (for creates) */
278b4556396SSage Weil 
2792f2dc053SSage Weil #define CEPH_MDS_R_DIRECT_IS_HASH	(1) /* r_direct_hash is valid */
2802f2dc053SSage Weil #define CEPH_MDS_R_ABORTED		(2) /* call was aborted */
281c36d6414SJeff Layton #define CEPH_MDS_R_GOT_UNSAFE		(3) /* got an unsafe reply */
2829eaa7b79SJeff Layton #define CEPH_MDS_R_GOT_SAFE		(4) /* got a safe reply */
2830ed1e90aSArnd Bergmann #define CEPH_MDS_R_GOT_RESULT		(5) /* got a result */
2842f2dc053SSage Weil #define CEPH_MDS_R_DID_PREPOPULATE	(6) /* prepopulated readdir */
2852f2dc053SSage Weil #define CEPH_MDS_R_PARENT_LOCKED	(7) /* is r_parent->i_rwsem wlocked? */
2862f2dc053SSage Weil #define CEPH_MDS_R_ASYNC		(8) /* async request */
2872f2dc053SSage Weil #define CEPH_MDS_R_FSCRYPT_FILE		(9) /* must marshal fscrypt_file field */
2882f2dc053SSage Weil 	unsigned long	r_req_flags;
2892f2dc053SSage Weil 
29025e6bae3SYan, Zheng 	struct mutex r_fill_mutex;
2912f2dc053SSage Weil 
2922f2dc053SSage Weil 	union ceph_mds_request_args r_args;
2932f2dc053SSage Weil 
2942f2dc053SSage Weil 	struct ceph_fscrypt_auth *r_fscrypt_auth;
2952f2dc053SSage Weil 	u64	r_fscrypt_file;
2962f2dc053SSage Weil 
2972f2dc053SSage Weil 	u8 *r_altname;		    /* fscrypt binary crypttext for long filenames */
2982f2dc053SSage Weil 	u32 r_altname_len;	    /* length of r_altname */
2992f2dc053SSage Weil 
3002f2dc053SSage Weil 	int r_fmode;        /* file mode, if expecting cap */
3012f2dc053SSage Weil 	int r_request_release_offset;
3022f2dc053SSage Weil 	const struct cred *r_cred;
3039eaa7b79SJeff Layton 	struct timespec64 r_stamp;
304a25949b9SJeff Layton 
305a25949b9SJeff Layton 	/* for choosing which mds to send this request to */
306a25949b9SJeff Layton 	int r_direct_mode;
307c36d6414SJeff Layton 	u32 r_direct_hash;      /* choose dir frag based on this dentry hash */
3082f2dc053SSage Weil 
309a319bf56SIlya Dryomov 	/* data payload is used for xattr ops */
3102f2dc053SSage Weil 	struct ceph_pagelist *r_pagelist;
31170c94820SXiubo Li 
31270c94820SXiubo Li 	/* what caps shall we drop? */
3132f2dc053SSage Weil 	int r_inode_drop, r_inode_unless;
3142f2dc053SSage Weil 	int r_dentry_drop, r_dentry_unless;
3152f2dc053SSage Weil 	int r_old_dentry_drop, r_old_dentry_unless;
3162f2dc053SSage Weil 	struct inode *r_old_inode;
3172f2dc053SSage Weil 	int r_old_inode_drop, r_old_inode_unless;
3182f2dc053SSage Weil 
3192f2dc053SSage Weil 	struct ceph_msg  *r_request;  /* original request */
32068cd5b4bSYan, Zheng 	struct ceph_msg  *r_reply;
32168cd5b4bSYan, Zheng 	struct ceph_mds_reply_info_parsed r_reply_info;
32268cd5b4bSYan, Zheng 	int r_err;
3232f2dc053SSage Weil 	u32               r_readdir_offset;
3242f2dc053SSage Weil 
3252f2dc053SSage Weil 	struct page *r_locked_page;
3262f2dc053SSage Weil 	int r_dir_caps;
3272f2dc053SSage Weil 	int r_num_caps;
328e55b71f8SGreg Farnum 
3296deb8008SJeff Layton 	unsigned long r_timeout;  /* optional.  jiffies, 0 is "wait forever" */
3302f2dc053SSage Weil 	unsigned long r_started;  /* start time to measure timeout against */
3312f2dc053SSage Weil 	unsigned long r_start_latency;  /* start time to measure latency */
3322f2dc053SSage Weil 	unsigned long r_end_latency;    /* finish time to measure latency */
3332f2dc053SSage Weil 	unsigned long r_request_started; /* start time for mds request only,
3342f2dc053SSage Weil 					    used to measure lease durations */
3352f2dc053SSage Weil 
3362f2dc053SSage Weil 	/* link unsafe requests to parent directory, for fsync */
337fdd4e158SYan, Zheng 	struct inode	*r_unsafe_dir;
338fdd4e158SYan, Zheng 	struct list_head r_unsafe_dir_item;
339fdd4e158SYan, Zheng 
3402f2dc053SSage Weil 	/* unsafe requests that modify the target inode */
341*6eb06c46SXiubo Li 	struct list_head r_unsafe_target_item;
342*6eb06c46SXiubo Li 
3432f2dc053SSage Weil 	struct ceph_mds_session *r_session;
3442f2dc053SSage Weil 
3452f2dc053SSage Weil 	int               r_attempts;   /* resend attempts */
34610183a69SYan, Zheng 	int               r_num_fwd;    /* number of forward attempts */
34710183a69SYan, Zheng 	int               r_resend_mds; /* mds to resend to next, if any*/
34810183a69SYan, Zheng 	u32               r_sent_on_mseq; /* cap mseq request was sent at*/
3497627151eSYan, Zheng 	u64		  r_deleg_ino;
350779fe0fbSYan, Zheng 
351779fe0fbSYan, Zheng 	struct list_head  r_wait;
35210183a69SYan, Zheng 	struct completion r_completion;
35310183a69SYan, Zheng 	struct completion r_safe_completion;
35475c9627eSYan, Zheng 	ceph_mds_request_callback_t r_callback;
35575c9627eSYan, Zheng 	struct list_head  r_unsafe_item;  /* per-session unsafe list item */
35675c9627eSYan, Zheng 
35775c9627eSYan, Zheng 	long long	  r_dir_release_cnt;
35875c9627eSYan, Zheng 	long long	  r_dir_ordered_cnt;
35975c9627eSYan, Zheng 	int		  r_readdir_cache_idx;
36075c9627eSYan, Zheng 
36175c9627eSYan, Zheng 	int		  r_feature_needed;
36275c9627eSYan, Zheng 
3632f2dc053SSage Weil 	struct ceph_cap_reservation r_caps_reservation;
3640c44a8e0SLuis Henriques };
3650c44a8e0SLuis Henriques 
3660c44a8e0SLuis Henriques struct ceph_pool_perm {
3670c44a8e0SLuis Henriques 	struct rb_node node;
3680c44a8e0SLuis Henriques 	int perm;
3690c44a8e0SLuis Henriques 	s64 pool;
3700c44a8e0SLuis Henriques 	size_t pool_ns_len;
3710c44a8e0SLuis Henriques 	char pool_ns[];
3720c44a8e0SLuis Henriques };
3730c44a8e0SLuis Henriques 
3740c44a8e0SLuis Henriques struct ceph_snapid_map {
3753a3430afSJeff Layton 	struct rb_node node;
3763a3430afSJeff Layton 	struct list_head lru;
377ebce3eb2SJeff Layton 	atomic_t ref;
3783a3430afSJeff Layton 	dev_t dev;
3793a3430afSJeff Layton 	u64 snap;
3803a3430afSJeff Layton 	unsigned long last_used;
3813a3430afSJeff Layton };
3823a3430afSJeff Layton 
3830c44a8e0SLuis Henriques /*
3842f2dc053SSage Weil  * node for list of quotarealm inodes that are not visible from the filesystem
3852f2dc053SSage Weil  * mountpoint, but required to handle, e.g. quotas.
3862f2dc053SSage Weil  */
3873d14c5d2SYehuda Sadeh struct ceph_quotarealm_inode {
3882f2dc053SSage Weil 	struct rb_node node;
3892f2dc053SSage Weil 	u64 ino;
3902f2dc053SSage Weil 	unsigned long timeout; /* last time a lookup failed for this inode */
391f3c60c59SSage Weil 	struct mutex mutex;
392f3c60c59SSage Weil 	struct inode *inode;
3932f2dc053SSage Weil };
394430afbadSYan, Zheng 
3952f2dc053SSage Weil struct cap_wait {
3962f2dc053SSage Weil 	struct list_head	list;
39786d8f67bSYan, Zheng 	u64			ino;
398c74d79afSJeff Layton 	pid_t			tgid;
3992f2dc053SSage Weil 	int			need;
4002f2dc053SSage Weil 	int			want;
401d557c48dSLuis Henriques };
4020c44a8e0SLuis Henriques 
4030c44a8e0SLuis Henriques enum {
4040c44a8e0SLuis Henriques 	CEPH_MDSC_STOPPING_BEGIN = 1,
4050c44a8e0SLuis Henriques 	CEPH_MDSC_STOPPING_FLUSHING = 2,
4060c44a8e0SLuis Henriques 	CEPH_MDSC_STOPPING_FLUSHED = 3,
4070c44a8e0SLuis Henriques };
408d557c48dSLuis Henriques 
4092f2dc053SSage Weil /*
4102f2dc053SSage Weil  * mds client state
4112f2dc053SSage Weil  */
4122f2dc053SSage Weil struct ceph_mds_client {
4132f2dc053SSage Weil 	struct ceph_fs_client  *fsc;
4142f2dc053SSage Weil 	struct mutex            mutex;         /* all nested structures */
4152f2dc053SSage Weil 
416affbc19aSYan, Zheng 	struct ceph_mdsmap      *mdsmap;
4172f2dc053SSage Weil 	struct completion       safe_umount_waiters;
418a105f00cSSage Weil 	wait_queue_head_t       session_close_wq;
4192f2dc053SSage Weil 	struct list_head        waiting_for_map;
42081c5a148SYan, Zheng 	int 			mdsmap_err;
4212f2dc053SSage Weil 
4222f2dc053SSage Weil 	struct ceph_mds_session **sessions;    /* NULL for mds if no session */
4232f2dc053SSage Weil 	atomic_t		num_sessions;
424e8a7b8b1SYan, Zheng 	int                     max_sessions;  /* len of sessions array */
425e8a7b8b1SYan, Zheng 
42644ca18f2SSage Weil 	spinlock_t              stopping_lock;  /* protect snap_empty */
4272f2dc053SSage Weil 	int                     stopping;      /* the stage of shutting down */
4282f2dc053SSage Weil 	atomic_t                stopping_blockers;
4292f2dc053SSage Weil 	struct completion	stopping_waiter;
4302f2dc053SSage Weil 
4312f2dc053SSage Weil 	atomic64_t		quotarealms_count; /* # realms with quota */
4322f2dc053SSage Weil 	/*
4332f2dc053SSage Weil 	 * We keep a list of inodes we don't see in the mountpoint but that we
434553adfd9SYan, Zheng 	 * need to track quota realms.
435e4500b5eSYan, Zheng 	 */
436db354052SSage Weil 	struct rb_root		quotarealms_inodes;
4372f2dc053SSage Weil 	struct mutex		quotarealms_inodes_mutex;
4382f2dc053SSage Weil 
4392f2dc053SSage Weil 	/*
4402f2dc053SSage Weil 	 * snap_rwsem will cover cap linkage into snaprealms, and
44137c4efc1SYan, Zheng 	 * realm snap contexts.  (later, we can do per-realm snap
442fe33032dSYan, Zheng 	 * contexts locks..)  the empty list contains realms with no
44337c4efc1SYan, Zheng 	 * references (implying they contain no inodes with caps) that
44437151668SYehuda Sadeh 	 * should be destroyed.
44537151668SYehuda Sadeh 	 */
44637151668SYehuda Sadeh 	u64			last_snap_seq;
44737151668SYehuda Sadeh 	struct rw_semaphore     snap_rwsem;
44837151668SYehuda Sadeh 	struct rb_root          snap_realms;
44937151668SYehuda Sadeh 	struct list_head        snap_empty;
45037151668SYehuda Sadeh 	int			num_snap_realms;
45137151668SYehuda Sadeh 	spinlock_t              snap_empty_lock;  /* protect snap_empty */
45237151668SYehuda Sadeh 
45337151668SYehuda Sadeh 	u64                    last_tid;      /* most recent mds request */
45437151668SYehuda Sadeh 	u64                    oldest_tid;    /* oldest incomplete mds request,
45537151668SYehuda Sadeh 						 excluding setfilelock requests */
45637151668SYehuda Sadeh 	struct rb_root         request_tree;  /* pending mds requests */
45737151668SYehuda Sadeh 	struct delayed_work    delayed_work;  /* delayed work */
4583a3430afSJeff Layton 	unsigned long    last_renew_caps;  /* last time we renewed our caps */
45937151668SYehuda Sadeh 	struct list_head cap_delay_list;   /* caps with delayed release */
46037151668SYehuda Sadeh 	spinlock_t       cap_delay_lock;   /* protects cap_delay_list */
461fe33032dSYan, Zheng 	struct list_head snap_flush_list;  /* cap_snaps ready to flush */
46237151668SYehuda Sadeh 	spinlock_t       snap_flush_lock;
46337151668SYehuda Sadeh 
46437151668SYehuda Sadeh 	u64               last_cap_flush_tid;
46537151668SYehuda Sadeh 	struct list_head  cap_flush_list;
46637c4efc1SYan, Zheng 	struct list_head  cap_dirty_migrating; /* ...that are migration... */
46737c4efc1SYan, Zheng 	int               num_cap_flushing; /* # caps we are flushing */
46837c4efc1SYan, Zheng 	spinlock_t        cap_dirty_lock;   /* protects above items */
46910183a69SYan, Zheng 	wait_queue_head_t cap_flushing_wq;
470f9009efaSXiubo Li 
471f9009efaSXiubo Li 	struct work_struct cap_reclaim_work;
47275c9627eSYan, Zheng 	atomic_t	   cap_reclaim_pending;
47375c9627eSYan, Zheng 
47475c9627eSYan, Zheng 	/*
47575c9627eSYan, Zheng 	 * Cap reservations
47610183a69SYan, Zheng 	 *
47710183a69SYan, Zheng 	 * Maintain a global pool of preallocated struct ceph_caps, referenced
478717e6f28SYan, Zheng 	 * by struct ceph_caps_reservations.  This ensures that we preallocate
479717e6f28SYan, Zheng 	 * memory needed to successfully process an MDS response.  (If an MDS
4802f2dc053SSage Weil 	 * sends us cap information and we fail to process it, we will have
4812f2dc053SSage Weil 	 * problems due to the client and MDS being out of sync.)
4822f2dc053SSage Weil 	 *
4832f2dc053SSage Weil 	 * Reservations are 'owned' by a ceph_cap_reservation context.
4843e699bd8SXiubo Li 	 */
48562575e27SJeff Layton 	spinlock_t	caps_list_lock;
4863e699bd8SXiubo Li 	struct		list_head caps_list; /* unused (reserved or
4872f2dc053SSage Weil 						unreserved) */
4882f2dc053SSage Weil 	struct		list_head cap_wait_list;
4892f2dc053SSage Weil 	int		caps_total_count;    /* total caps allocated */
490a687ecafSJohn Spray 	int		caps_use_count;      /* in use */
491a687ecafSJohn Spray 	int		caps_use_max;	     /* max used caps */
4925b3248c6SXiubo Li 	int		caps_reserve_count;  /* unused, reserved */
4935b3248c6SXiubo Li 	int		caps_avail_count;    /* unused, unreserved */
4942f2dc053SSage Weil 	int		caps_min_count;      /* keep at least this many
4952f2dc053SSage Weil 						(unreserved) */
4962f2dc053SSage Weil 	spinlock_t	  dentry_list_lock;
4972f2dc053SSage Weil 	struct list_head  dentry_leases;     /* fifo list */
4982f2dc053SSage Weil 	struct list_head  dentry_dir_leases; /* lru list */
4993d14c5d2SYehuda Sadeh 
5002f2dc053SSage Weil 	struct ceph_client_metric metric;
50148fec5d0SYan, Zheng 
5023d14c5d2SYehuda Sadeh 	spinlock_t		snapid_map_lock;
5032f2dc053SSage Weil 	struct rb_root		snapid_map_tree;
5042f2dc053SSage Weil 	struct list_head	snapid_map_lru;
5052f2dc053SSage Weil 
506167c9e35SSage Weil 	struct rw_semaphore     pool_perm_rwsem;
50754008399SYan, Zheng 	struct rb_root		pool_perm_tree;
50854008399SYan, Zheng 
5092f2dc053SSage Weil 	char nodename[__NEW_UTS_LEN + 1];
5102f2dc053SSage Weil };
51186bda539SJeff Layton 
51286bda539SJeff Layton extern const char *ceph_mds_op_name(int op);
5132f2dc053SSage Weil 
5149eaa7b79SJeff Layton extern bool check_session_state(struct ceph_mds_session *s);
5159eaa7b79SJeff Layton void inc_session_sequence(struct ceph_mds_session *s);
5169eaa7b79SJeff Layton 
5172f2dc053SSage Weil extern struct ceph_mds_session *
5182f2dc053SSage Weil __ceph_lookup_mds_session(struct ceph_mds_client *, int mds);
5192f2dc053SSage Weil 
520a25949b9SJeff Layton extern const char *ceph_session_state_name(int s);
521e64f44a8SXiubo Li 
5222f2dc053SSage Weil extern struct ceph_mds_session *
5232f2dc053SSage Weil ceph_get_mds_session(struct ceph_mds_session *s);
524153c8e6bSSage Weil extern void ceph_put_mds_session(struct ceph_mds_session *s);
5252f2dc053SSage Weil 
526153c8e6bSSage Weil extern int ceph_send_msg_mds(struct ceph_mds_client *mdsc,
527153c8e6bSSage Weil 			     struct ceph_msg *msg, int mds);
528153c8e6bSSage Weil 
529153c8e6bSSage Weil extern int ceph_mdsc_init(struct ceph_fs_client *fsc);
530153c8e6bSSage Weil extern void ceph_mdsc_close_sessions(struct ceph_mds_client *mdsc);
5312f2dc053SSage Weil extern void ceph_mdsc_force_umount(struct ceph_mds_client *mdsc);
532d095559cSXiubo Li extern void ceph_mdsc_destroy(struct ceph_fs_client *fsc);
53359b312f3SXiubo Li 
53459b312f3SXiubo Li extern void ceph_mdsc_sync(struct ceph_mds_client *mdsc);
53559b312f3SXiubo Li 
536fba97e80SXiubo Li extern void ceph_invalidate_dir_request(struct ceph_mds_request *req);
537e3ec8d68SYan, Zheng extern int ceph_alloc_readdir_reply_buffer(struct ceph_mds_request *req,
538e3ec8d68SYan, Zheng 					   struct inode *dir);
539e3ec8d68SYan, Zheng extern struct ceph_mds_request *
5403d7ded4dSSage Weil ceph_mdsc_create_request(struct ceph_mds_client *mdsc, int op, int mode);
54137c4efc1SYan, Zheng extern int ceph_mdsc_submit_request(struct ceph_mds_client *mdsc,
542fe33032dSYan, Zheng 				    struct inode *dir,
543f5d77269SJeff Layton 				    struct ceph_mds_request *req);
544f5d77269SJeff Layton int ceph_mdsc_wait_request(struct ceph_mds_client *mdsc,
545f5d77269SJeff Layton 			struct ceph_mds_request *req,
546f5d77269SJeff Layton 			ceph_mds_request_wait_callback_t wait_func);
5472f2dc053SSage Weil extern int ceph_mdsc_do_request(struct ceph_mds_client *mdsc,
5482f2dc053SSage Weil 				struct inode *dir,
549f77f21bbSJeff Layton 				struct ceph_mds_request *req);
550f77f21bbSJeff Layton extern void ceph_mdsc_release_dir_caps(struct ceph_mds_request *req);
5512a575f13SJeff Layton extern void ceph_mdsc_release_dir_caps_no_check(struct ceph_mds_request *req);
ceph_mdsc_get_request(struct ceph_mds_request * req)552f77f21bbSJeff Layton static inline void ceph_mdsc_get_request(struct ceph_mds_request *req)
553f77f21bbSJeff Layton {
554f77f21bbSJeff Layton 	kref_get(&req->r_kref);
5552f2dc053SSage Weil }
5562f2dc053SSage Weil extern void ceph_mdsc_release_request(struct kref *kref);
ceph_mdsc_put_request(struct ceph_mds_request * req)5572f2dc053SSage Weil static inline void ceph_mdsc_put_request(struct ceph_mds_request *req)
5582f2dc053SSage Weil {
5592f2dc053SSage Weil 	kref_put(&req->r_kref, ceph_mdsc_release_request);
5602f2dc053SSage Weil }
5612f2dc053SSage Weil 
5622f2dc053SSage Weil extern void send_flush_mdlog(struct ceph_mds_session *s);
563430afbadSYan, Zheng extern void ceph_mdsc_iterate_sessions(struct ceph_mds_client *mdsc,
564430afbadSYan, Zheng 				       void (*cb)(struct ceph_mds_session *),
565430afbadSYan, Zheng 				       bool check_state);
5662f2dc053SSage Weil extern struct ceph_msg *ceph_create_session_msg(u32 op, u64 seq);
5672f2dc053SSage Weil extern void __ceph_queue_cap_release(struct ceph_mds_session *session,
5685d72d13cSYan, Zheng 				    struct ceph_cap *cap);
5695d72d13cSYan, Zheng extern void ceph_flush_cap_releases(struct ceph_mds_client *mdsc,
570154f42c2SSage Weil 				    struct ceph_mds_session *session);
571154f42c2SSage Weil extern void ceph_queue_cap_reclaim_work(struct ceph_mds_client *mdsc);
572154f42c2SSage Weil extern void ceph_reclaim_caps_nr(struct ceph_mds_client *mdsc, int nr);
573e30ee581SZhi Zhang extern int ceph_iterate_session_caps(struct ceph_mds_session *session,
574e30ee581SZhi Zhang 				     int (*cb)(struct inode *, int mds, void *),
575e30ee581SZhi Zhang 				     void *arg);
576d4846487SJeff Layton extern void ceph_mdsc_pre_umount(struct ceph_mds_client *mdsc);
577891f3f5aSJeff Layton 
ceph_mdsc_free_path(char * path,int len)578891f3f5aSJeff Layton static inline void ceph_mdsc_free_path(char *path, int len)
579891f3f5aSJeff Layton {
580891f3f5aSJeff Layton 	if (!IS_ERR_OR_NULL(path))
581891f3f5aSJeff Layton 		__putname(path - (PATH_MAX - 1 - len));
5825e56776dSXiubo Li }
583891f3f5aSJeff Layton 
584d4846487SJeff Layton extern char *ceph_mdsc_build_path(struct ceph_mds_client *mdsc,
5854868e537SXiubo Li 				  struct dentry *dentry, int *plen, u64 *base,
586d4846487SJeff Layton 				  int for_wire);
587d4846487SJeff Layton 
5882f2dc053SSage Weil extern void __ceph_mdsc_drop_dentry_lease(struct dentry *dentry);
589 extern void ceph_mdsc_lease_send_msg(struct ceph_mds_session *session,
590 				     struct dentry *dentry, char action,
591 				     u32 seq);
592 
593 extern void ceph_mdsc_handle_mdsmap(struct ceph_mds_client *mdsc,
594 				    struct ceph_msg *msg);
595 extern void ceph_mdsc_handle_fsmap(struct ceph_mds_client *mdsc,
596 				   struct ceph_msg *msg);
597 
598 extern struct ceph_mds_session *
599 ceph_mdsc_open_export_target_session(struct ceph_mds_client *mdsc, int target);
600 extern void ceph_mdsc_open_export_target_sessions(struct ceph_mds_client *mdsc,
601 					  struct ceph_mds_session *session);
602 
603 extern int ceph_trim_caps(struct ceph_mds_client *mdsc,
604 			  struct ceph_mds_session *session,
605 			  int max_caps);
606 
ceph_wait_on_async_create(struct inode * inode)607 static inline int ceph_wait_on_async_create(struct inode *inode)
608 {
609 	struct ceph_inode_info *ci = ceph_inode(inode);
610 
611 	return wait_on_bit(&ci->i_ceph_flags, CEPH_ASYNC_CREATE_BIT,
612 			   TASK_KILLABLE);
613 }
614 
615 extern int ceph_wait_on_conflict_unlink(struct dentry *dentry);
616 extern u64 ceph_get_deleg_ino(struct ceph_mds_session *session);
617 extern int ceph_restore_deleg_ino(struct ceph_mds_session *session, u64 ino);
618 #endif
619