[linux-2.6-block.git] / include / linux / ceph / osd_client.h

#ifndef _FS_CEPH_OSD_CLIENT_H
#define _FS_CEPH_OSD_CLIENT_H

#include <linux/completion.h>
#include <linux/kref.h>
#include <linux/mempool.h>
#include <linux/rbtree.h>

#include "types.h"
#include "osdmap.h"
#include "messenger.h"

/* 
 * Maximum object name size 
 * (must be at least as big as RBD_MAX_MD_NAME_LEN -- currently 100) 
 */
#define MAX_OBJ_NAME_SIZE 100

struct ceph_msg;
struct ceph_snap_context;
struct ceph_osd_request;
struct ceph_osd_client;
struct ceph_authorizer;
struct ceph_pagelist;

/*
 * completion callback for async writepages
 */
typedef void (*ceph_osdc_callback_t)(struct ceph_osd_request *,
				     struct ceph_msg *);

/* a given osd we're communicating with */
struct ceph_osd {
	atomic_t o_ref;
	struct ceph_osd_client *o_osdc;
	int o_osd;
	int o_incarnation;
	struct rb_node o_node;
	struct ceph_connection o_con;
	struct list_head o_requests;
	struct list_head o_linger_requests;
	struct list_head o_osd_lru;
	struct ceph_authorizer *o_authorizer;
	void *o_authorizer_buf, *o_authorizer_reply_buf;
	size_t o_authorizer_buf_len, o_authorizer_reply_buf_len;
	unsigned long lru_ttl;
	int o_marked_for_keepalive;
	struct list_head o_keepalive_item;
};

/* an in-flight request */
struct ceph_osd_request {
	u64             r_tid;              /* unique for this client */
	struct rb_node  r_node;
	struct list_head r_req_lru_item;
	struct list_head r_osd_item;
	struct list_head r_linger_item;
	struct list_head r_linger_osd;
	struct ceph_osd *r_osd;
	struct ceph_pg   r_pgid;
	int              r_pg_osds[CEPH_PG_MAX_SIZE];
	int              r_num_pg_osds;

	struct ceph_connection *r_con_filling_msg;

	struct ceph_msg  *r_request, *r_reply;
	int               r_result;
	int               r_flags;     /* any additional flags for the osd */
	u32               r_sent;      /* >0 if r_request is sending/sent */
	int               r_got_reply;
	int		  r_linger;

	struct ceph_osd_client *r_osdc;
	struct kref       r_kref;
	bool              r_mempool;
	struct completion r_completion, r_safe_completion;
	ceph_osdc_callback_t r_callback, r_safe_callback;
	struct ceph_eversion r_reassert_version;
	struct list_head  r_unsafe_item;

	struct inode *r_inode;         	      /* for use by callbacks */
	void *r_priv;			      /* ditto */

	char              r_oid[MAX_OBJ_NAME_SIZE];          /* object name */
	int               r_oid_len;
	unsigned long     r_stamp;            /* send OR check time */

	struct ceph_file_layout r_file_layout;
	struct ceph_snap_context *r_snapc;    /* snap context for writes */
	unsigned          r_num_pages;        /* size of page array (follows) */
	unsigned          r_page_alignment;   /* io offset in first page */
	struct page     **r_pages;            /* pages for data payload */
	int               r_pages_from_pool;
	int               r_own_pages;        /* if true, i own page list */
#ifdef CONFIG_BLOCK
	struct bio       *r_bio;	      /* instead of pages */
#endif

	struct ceph_pagelist *r_trail;	      /* trailing part of the data */
};

struct ceph_osd_event {
	u64 cookie;
	int one_shot;
	struct ceph_osd_client *osdc;
	void (*cb)(u64, u64, u8, void *);
	void *data;
	struct rb_node node;
	struct list_head osd_node;
	struct kref kref;
	struct completion completion;
};

struct ceph_osd_event_work {
	struct work_struct work;
	struct ceph_osd_event *event;
        u64 ver;
        u64 notify_id;
        u8 opcode;
};

struct ceph_osd_client {
	struct ceph_client     *client;

	struct ceph_osdmap     *osdmap;       /* current map */
	struct rw_semaphore    map_sem;
	struct completion      map_waiters;
	u64                    last_requested_map;

	struct mutex           request_mutex;
	struct rb_root         osds;          /* osds */
	struct list_head       osd_lru;       /* idle osds */
	u64                    timeout_tid;   /* tid of timeout triggering rq */
	u64                    last_tid;      /* tid of last request */
	struct rb_root         requests;      /* pending requests */
	struct list_head       req_lru;	      /* in-flight lru */
	struct list_head       req_unsent;    /* unsent/need-resend queue */
	struct list_head       req_notarget;  /* map to no osd */
	struct list_head       req_linger;    /* lingering requests */
	int                    num_requests;
	struct delayed_work    timeout_work;
	struct delayed_work    osds_timeout_work;
#ifdef CONFIG_DEBUG_FS
	struct dentry 	       *debugfs_file;
#endif

	mempool_t              *req_mempool;

	struct ceph_msgpool	msgpool_op;
	struct ceph_msgpool	msgpool_op_reply;

	spinlock_t		event_lock;
	struct rb_root		event_tree;
	u64			event_count;

	struct workqueue_struct	*notify_wq;
};

struct ceph_osd_req_op {
	u16 op;           /* CEPH_OSD_OP_* */
	u32 flags;        /* CEPH_OSD_FLAG_* */
	union {
		struct {
			u64 offset, length;
			u64 truncate_size;
			u32 truncate_seq;
		} extent;
		struct {
			const char *name;
			u32 name_len;
			const char  *val;
			u32 value_len;
			__u8 cmp_op;       /* CEPH_OSD_CMPXATTR_OP_* */
			__u8 cmp_mode;     /* CEPH_OSD_CMPXATTR_MODE_* */
		} xattr;
		struct {
			const char *class_name;
			__u8 class_len;
			const char *method_name;
			__u8 method_len;
			__u8 argc;
			const char *indata;
			u32 indata_len;
		} cls;
		struct {
			u64 cookie, count;
		} pgls;
	        struct {
		        u64 snapid;
	        } snap;
		struct {
			u64 cookie;
			u64 ver;
			__u8 flag;
			u32 prot_ver;
			u32 timeout;
		} watch;
	};
	u32 payload_len;
};

extern int ceph_osdc_init(struct ceph_osd_client *osdc,
			  struct ceph_client *client);
extern void ceph_osdc_stop(struct ceph_osd_client *osdc);

extern void ceph_osdc_handle_reply(struct ceph_osd_client *osdc,
				   struct ceph_msg *msg);
extern void ceph_osdc_handle_map(struct ceph_osd_client *osdc,
				 struct ceph_msg *msg);

extern void ceph_calc_raw_layout(struct ceph_osd_client *osdc,
			struct ceph_file_layout *layout,
			u64 snapid,
			u64 off, u64 *plen, u64 *bno,
			struct ceph_osd_request *req,
			struct ceph_osd_req_op *op);

extern struct ceph_osd_request *ceph_osdc_alloc_request(struct ceph_osd_client *osdc,
					       int flags,
					       struct ceph_snap_context *snapc,
					       struct ceph_osd_req_op *ops,
					       bool use_mempool,
					       gfp_t gfp_flags,
					       struct page **pages,
					       struct bio *bio);

extern void ceph_osdc_build_request(struct ceph_osd_request *req,
				    u64 off, u64 *plen,
				    struct ceph_osd_req_op *src_ops,
				    struct ceph_snap_context *snapc,
				    struct timespec *mtime,
				    const char *oid,
				    int oid_len);

extern struct ceph_osd_request *ceph_osdc_new_request(struct ceph_osd_client *,
				      struct ceph_file_layout *layout,
				      struct ceph_vino vino,
				      u64 offset, u64 *len, int op, int flags,
				      struct ceph_snap_context *snapc,
				      int do_sync, u32 truncate_seq,
				      u64 truncate_size,
				      struct timespec *mtime,
				      bool use_mempool, int num_reply,
				      int page_align);

extern void ceph_osdc_set_request_linger(struct ceph_osd_client *osdc,
					 struct ceph_osd_request *req);
extern void ceph_osdc_unregister_linger_request(struct ceph_osd_client *osdc,
						struct ceph_osd_request *req);

static inline void ceph_osdc_get_request(struct ceph_osd_request *req)
{
	kref_get(&req->r_kref);
}
extern void ceph_osdc_release_request(struct kref *kref);
static inline void ceph_osdc_put_request(struct ceph_osd_request *req)
{
	kref_put(&req->r_kref, ceph_osdc_release_request);
}

extern int ceph_osdc_start_request(struct ceph_osd_client *osdc,
				   struct ceph_osd_request *req,
				   bool nofail);
extern int ceph_osdc_wait_request(struct ceph_osd_client *osdc,
				  struct ceph_osd_request *req);
extern void ceph_osdc_sync(struct ceph_osd_client *osdc);

extern int ceph_osdc_readpages(struct ceph_osd_client *osdc,
			       struct ceph_vino vino,
			       struct ceph_file_layout *layout,
			       u64 off, u64 *plen,
			       u32 truncate_seq, u64 truncate_size,
			       struct page **pages, int nr_pages,
			       int page_align);

extern int ceph_osdc_writepages(struct ceph_osd_client *osdc,
				struct ceph_vino vino,
				struct ceph_file_layout *layout,
				struct ceph_snap_context *sc,
				u64 off, u64 len,
				u32 truncate_seq, u64 truncate_size,
				struct timespec *mtime,
				struct page **pages, int nr_pages,
				int flags, int do_sync, bool nofail);

/* watch/notify events */
extern int ceph_osdc_create_event(struct ceph_osd_client *osdc,
				  void (*event_cb)(u64, u64, u8, void *),
				  int one_shot, void *data,
				  struct ceph_osd_event **pevent);
extern void ceph_osdc_cancel_event(struct ceph_osd_event *event);
extern int ceph_osdc_wait_event(struct ceph_osd_event *event,
				unsigned long timeout);
extern void ceph_osdc_put_event(struct ceph_osd_event *event);
#endif
Commit	Line	Data
f24e9980 SW	1	#ifndef _FS_CEPH_OSD_CLIENT_H
	2	#define _FS_CEPH_OSD_CLIENT_H
	3
	4	#include <linux/completion.h>
415e49a9	5	#include <linux/kref.h>
f24e9980 SW	6	#include <linux/mempool.h>
	7	#include <linux/rbtree.h>
	8
	9	#include "types.h"
	10	#include "osdmap.h"
	11	#include "messenger.h"
	12
224736d9 SP	13	/*
	14	* Maximum object name size
	15	* (must be at least as big as RBD_MAX_MD_NAME_LEN -- currently 100)
	16	*/
	17	#define MAX_OBJ_NAME_SIZE 100
	18
f24e9980 SW	19	struct ceph_msg;
	20	struct ceph_snap_context;
	21	struct ceph_osd_request;
	22	struct ceph_osd_client;
4e7a5dcd	23	struct ceph_authorizer;
68b4476b	24	struct ceph_pagelist;
f24e9980 SW	25
	26	/*
	27	* completion callback for async writepages
	28	*/
	29	typedef void (ceph_osdc_callback_t)(struct ceph_osd_request ,
	30	struct ceph_msg *);
	31
	32	/* a given osd we're communicating with */
	33	struct ceph_osd {
	34	atomic_t o_ref;
	35	struct ceph_osd_client *o_osdc;
	36	int o_osd;
	37	int o_incarnation;
	38	struct rb_node o_node;
	39	struct ceph_connection o_con;
	40	struct list_head o_requests;
a40c4f10	41	struct list_head o_linger_requests;
f5a2041b	42	struct list_head o_osd_lru;
4e7a5dcd SW	43	struct ceph_authorizer *o_authorizer;
	44	void o_authorizer_buf, o_authorizer_reply_buf;
	45	size_t o_authorizer_buf_len, o_authorizer_reply_buf_len;
f5a2041b	46	unsigned long lru_ttl;
422d2cb8 YS	47	int o_marked_for_keepalive;
422d2cb8 YS	48	struct list_head o_keepalive_item;
f24e9980 SW	49	};
	50
	51	/* an in-flight request */
	52	struct ceph_osd_request {
	53	u64 r_tid; /* unique for this client */
	54	struct rb_node r_node;
422d2cb8	55	struct list_head r_req_lru_item;
f24e9980	56	struct list_head r_osd_item;
a40c4f10 YS	57	struct list_head r_linger_item;
a40c4f10 YS	58	struct list_head r_linger_osd;
f24e9980	59	struct ceph_osd *r_osd;
7740a42f	60	struct ceph_pg r_pgid;
d85b7056 SW	61	int r_pg_osds[CEPH_PG_MAX_SIZE];
d85b7056 SW	62	int r_num_pg_osds;
f24e9980	63
0d59ab81	64	struct ceph_connection *r_con_filling_msg;
350b1c32	65
f24e9980 SW	66	struct ceph_msg r_request, r_reply;
	67	int r_result;
	68	int r_flags; /* any additional flags for the osd */
	69	u32 r_sent; /* >0 if r_request is sending/sent */
350b1c32	70	int r_got_reply;
a40c4f10	71	int r_linger;
f24e9980 SW	72
f24e9980 SW	73	struct ceph_osd_client *r_osdc;
415e49a9	74	struct kref r_kref;
f24e9980 SW	75	bool r_mempool;
	76	struct completion r_completion, r_safe_completion;
	77	ceph_osdc_callback_t r_callback, r_safe_callback;
	78	struct ceph_eversion r_reassert_version;
	79	struct list_head r_unsafe_item;
	80
	81	struct inode r_inode; / for use by callbacks */
3d14c5d2	82	void r_priv; / ditto */
f24e9980	83
224736d9	84	char r_oid[MAX_OBJ_NAME_SIZE]; /* object name */
f24e9980	85	int r_oid_len;
3dd72fc0	86	unsigned long r_stamp; /* send OR check time */
f24e9980 SW	87
	88	struct ceph_file_layout r_file_layout;
	89	struct ceph_snap_context r_snapc; / snap context for writes */
	90	unsigned r_num_pages; /* size of page array (follows) */
b7495fc2	91	unsigned r_page_alignment; /* io offset in first page */
f24e9980 SW	92	struct page *r_pages; / pages for data payload */
	93	int r_pages_from_pool;
	94	int r_own_pages; /* if true, i own page list */
68b4476b YS	95	#ifdef CONFIG_BLOCK
	96	struct bio r_bio; / instead of pages */
	97	#endif
	98
	99	struct ceph_pagelist r_trail; / trailing part of the data */
f24e9980 SW	100	};
f24e9980 SW	101
a40c4f10 YS	102	struct ceph_osd_event {
	103	u64 cookie;
	104	int one_shot;
	105	struct ceph_osd_client *osdc;
	106	void (cb)(u64, u64, u8, void );
	107	void *data;
	108	struct rb_node node;
	109	struct list_head osd_node;
	110	struct kref kref;
	111	struct completion completion;
	112	};
	113
	114	struct ceph_osd_event_work {
	115	struct work_struct work;
	116	struct ceph_osd_event *event;
	117	u64 ver;
	118	u64 notify_id;
	119	u8 opcode;
	120	};
	121
f24e9980 SW	122	struct ceph_osd_client {
	123	struct ceph_client *client;
	124
	125	struct ceph_osdmap osdmap; / current map */
	126	struct rw_semaphore map_sem;
	127	struct completion map_waiters;
	128	u64 last_requested_map;
	129
	130	struct mutex request_mutex;
	131	struct rb_root osds; /* osds */
f5a2041b	132	struct list_head osd_lru; /* idle osds */
f24e9980 SW	133	u64 timeout_tid; /* tid of timeout triggering rq */
	134	u64 last_tid; /* tid of last request */
	135	struct rb_root requests; /* pending requests */
6f6c7006 SW	136	struct list_head req_lru; /* in-flight lru */
	137	struct list_head req_unsent; /* unsent/need-resend queue */
	138	struct list_head req_notarget; /* map to no osd */
a40c4f10	139	struct list_head req_linger; /* lingering requests */
f24e9980 SW	140	int num_requests;
f24e9980 SW	141	struct delayed_work timeout_work;
f5a2041b	142	struct delayed_work osds_timeout_work;
039934b8	143	#ifdef CONFIG_DEBUG_FS
f24e9980	144	struct dentry *debugfs_file;
039934b8	145	#endif
f24e9980 SW	146
	147	mempool_t *req_mempool;
	148
0d59ab81	149	struct ceph_msgpool msgpool_op;
c16e7869	150	struct ceph_msgpool msgpool_op_reply;
a40c4f10 YS	151
	152	spinlock_t event_lock;
	153	struct rb_root event_tree;
	154	u64 event_count;
	155
	156	struct workqueue_struct *notify_wq;
f24e9980 SW	157	};
f24e9980 SW	158
68b4476b YS	159	struct ceph_osd_req_op {
	160	u16 op; /* CEPH_OSD_OP_* */
	161	u32 flags; /* CEPH_OSD_FLAG_* */
	162	union {
	163	struct {
	164	u64 offset, length;
	165	u64 truncate_size;
	166	u32 truncate_seq;
	167	} extent;
	168	struct {
	169	const char *name;
	170	u32 name_len;
	171	const char *val;
	172	u32 value_len;
	173	__u8 cmp_op; /* CEPH_OSD_CMPXATTR_OP_* */
	174	__u8 cmp_mode; /* CEPH_OSD_CMPXATTR_MODE_* */
	175	} xattr;
	176	struct {
ae1533b6	177	const char *class_name;
68b4476b	178	__u8 class_len;
ae1533b6	179	const char *method_name;
68b4476b YS	180	__u8 method_len;
68b4476b YS	181	__u8 argc;
ae1533b6	182	const char *indata;
68b4476b YS	183	u32 indata_len;
	184	} cls;
	185	struct {
	186	u64 cookie, count;
	187	} pgls;
ae1533b6 YS	188	struct {
	189	u64 snapid;
	190	} snap;
a40c4f10 YS	191	struct {
	192	u64 cookie;
	193	u64 ver;
	194	__u8 flag;
	195	u32 prot_ver;
	196	u32 timeout;
	197	} watch;
68b4476b YS	198	};
	199	u32 payload_len;
	200	};
	201
f24e9980 SW	202	extern int ceph_osdc_init(struct ceph_osd_client *osdc,
	203	struct ceph_client *client);
	204	extern void ceph_osdc_stop(struct ceph_osd_client *osdc);
	205
	206	extern void ceph_osdc_handle_reply(struct ceph_osd_client *osdc,
	207	struct ceph_msg *msg);
	208	extern void ceph_osdc_handle_map(struct ceph_osd_client *osdc,
	209	struct ceph_msg *msg);
	210
3499e8a5 YS	211	extern void ceph_calc_raw_layout(struct ceph_osd_client *osdc,
	212	struct ceph_file_layout *layout,
	213	u64 snapid,
68b4476b YS	214	u64 off, u64 plen, u64 bno,
	215	struct ceph_osd_request *req,
	216	struct ceph_osd_req_op *op);
3499e8a5 YS	217
	218	extern struct ceph_osd_request ceph_osdc_alloc_request(struct ceph_osd_client osdc,
	219	int flags,
	220	struct ceph_snap_context *snapc,
68b4476b	221	struct ceph_osd_req_op *ops,
3499e8a5 YS	222	bool use_mempool,
3499e8a5 YS	223	gfp_t gfp_flags,
68b4476b YS	224	struct page **pages,
68b4476b YS	225	struct bio *bio);
3499e8a5 YS	226
3499e8a5 YS	227	extern void ceph_osdc_build_request(struct ceph_osd_request *req,
68b4476b YS	228	u64 off, u64 *plen,
	229	struct ceph_osd_req_op *src_ops,
	230	struct ceph_snap_context *snapc,
	231	struct timespec *mtime,
	232	const char *oid,
	233	int oid_len);
3499e8a5	234
f24e9980 SW	235	extern struct ceph_osd_request ceph_osdc_new_request(struct ceph_osd_client ,
	236	struct ceph_file_layout *layout,
	237	struct ceph_vino vino,
	238	u64 offset, u64 *len, int op, int flags,
	239	struct ceph_snap_context *snapc,
	240	int do_sync, u32 truncate_seq,
	241	u64 truncate_size,
	242	struct timespec *mtime,
b7495fc2 SW	243	bool use_mempool, int num_reply,
b7495fc2 SW	244	int page_align);
f24e9980	245
a40c4f10 YS	246	extern void ceph_osdc_set_request_linger(struct ceph_osd_client *osdc,
	247	struct ceph_osd_request *req);
	248	extern void ceph_osdc_unregister_linger_request(struct ceph_osd_client *osdc,
	249	struct ceph_osd_request *req);
	250
f24e9980 SW	251	static inline void ceph_osdc_get_request(struct ceph_osd_request *req)
f24e9980 SW	252	{
415e49a9 SW	253	kref_get(&req->r_kref);
	254	}
	255	extern void ceph_osdc_release_request(struct kref *kref);
	256	static inline void ceph_osdc_put_request(struct ceph_osd_request *req)
	257	{
	258	kref_put(&req->r_kref, ceph_osdc_release_request);
f24e9980	259	}
f24e9980 SW	260
	261	extern int ceph_osdc_start_request(struct ceph_osd_client *osdc,
	262	struct ceph_osd_request *req,
	263	bool nofail);
	264	extern int ceph_osdc_wait_request(struct ceph_osd_client *osdc,
	265	struct ceph_osd_request *req);
	266	extern void ceph_osdc_sync(struct ceph_osd_client *osdc);
	267
	268	extern int ceph_osdc_readpages(struct ceph_osd_client *osdc,
	269	struct ceph_vino vino,
	270	struct ceph_file_layout *layout,
	271	u64 off, u64 *plen,
	272	u32 truncate_seq, u64 truncate_size,
b7495fc2 SW	273	struct page **pages, int nr_pages,
b7495fc2 SW	274	int page_align);
f24e9980 SW	275
	276	extern int ceph_osdc_writepages(struct ceph_osd_client *osdc,
	277	struct ceph_vino vino,
	278	struct ceph_file_layout *layout,
	279	struct ceph_snap_context *sc,
	280	u64 off, u64 len,
	281	u32 truncate_seq, u64 truncate_size,
	282	struct timespec *mtime,
	283	struct page **pages, int nr_pages,
	284	int flags, int do_sync, bool nofail);
	285
a40c4f10 YS	286	/* watch/notify events */
	287	extern int ceph_osdc_create_event(struct ceph_osd_client *osdc,
	288	void (event_cb)(u64, u64, u8, void ),
	289	int one_shot, void *data,
	290	struct ceph_osd_event **pevent);
	291	extern void ceph_osdc_cancel_event(struct ceph_osd_event *event);
	292	extern int ceph_osdc_wait_event(struct ceph_osd_event *event,
	293	unsigned long timeout);
	294	extern void ceph_osdc_put_event(struct ceph_osd_event *event);
f24e9980 SW	295	#endif
f24e9980 SW	296