X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Finclude%2Flu_target.h;h=e16e5bc52182bac30e0a8c3ace8021a97454b9fd;hp=929a4d43c941f283c04e85a836ff664cf164ef29;hb=c438fba7f068b0713d96dce1f0183ec6da7ab000;hpb=386818f0c56e438779e17d0ca12b481f17c53682

diff --git a/lustre/include/lu_target.h b/lustre/include/lu_target.h
index 929a4d4..e16e5bc 100644
--- a/lustre/include/lu_target.h
+++ b/lustre/include/lu_target.h
@@ -15,11 +15,7 @@
  *
  * You should have received a copy of the GNU General Public License
  * version 2 along with this program; If not, see
- * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
- *
- * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
- * CA 95054 USA or visit www.sun.com if you need additional information or
- * have any questions.
+ * http://www.gnu.org/licenses/gpl-2.0.html
  *
  * GPL HEADER END
  */
@@ -27,7 +23,7 @@
  * Copyright (c) 2009, 2010, Oracle and/or its affiliates. All rights reserved.
  * Use is subject to license terms.
  *
- * Copyright (c) 2011, 2012, Intel Corporation.
+ * Copyright (c) 2011, 2017, Intel Corporation.
  */
 /*
  * This file is part of Lustre, http://www.lustre.org/
@@ -43,9 +39,125 @@
 #include <lustre_disk.h>
 #include <lustre_lfsck.h>
 
+/* Each one represents a distribute transaction replay
+ * operation, and updates on each MDTs are linked to
+ * dtr_sub_list */
+struct distribute_txn_replay_req {
+	/* update record, may be vmalloc'd */
+	struct llog_update_record *dtrq_lur;
+	int			dtrq_lur_size;
+
+	/* linked to the distribute transaction replay
+	 * list (tdtd_replay_list) */
+	struct list_head	dtrq_list;
+	__u64			dtrq_master_transno;
+	__u64			dtrq_batchid;
+	__u64			dtrq_xid;
+
+	/* all of sub updates are linked here */
+	struct list_head	dtrq_sub_list;
+	spinlock_t		dtrq_sub_list_lock;
+
+	/* If the local update has been executed during replay */
+	__u32			dtrq_local_update_executed:1;
+};
+
+/* Each one represents a sub replay item under a distribute
+ * transaction. A distribute transaction will be operated in
+ * two or more MDTs, and updates on each MDT will be represented
+ * by this structure */
+struct distribute_txn_replay_req_sub {
+	__u32			dtrqs_mdt_index;
+
+	/* All of cookies for the update will be linked here */
+	spinlock_t		dtrqs_cookie_list_lock;
+	struct list_head	dtrqs_cookie_list;
+	struct list_head	dtrqs_list;
+};
+
+struct target_distribute_txn_data;
+typedef int (*distribute_txn_replay_handler_t)(struct lu_env *env,
+				       struct target_distribute_txn_data *tdtd,
+				       struct distribute_txn_replay_req *dtrq);
+typedef char *(*target_show_update_logs_retrievers_t)(void *data, int *size,
+						      int *count);
+struct target_distribute_txn_data {
+	/* Distribution ID is used to identify updates log on different
+	 * MDTs for one operation */
+	spinlock_t		tdtd_batchid_lock;
+	__u64			tdtd_batchid;
+	struct lu_target	*tdtd_lut;
+	struct dt_object	*tdtd_batchid_obj;
+	struct dt_device	*tdtd_dt;
+
+	/* Committed batchid for distribute transaction */
+	__u64                   tdtd_committed_batchid;
+
+	/* List for distribute transaction */
+	struct list_head	tdtd_list;
+
+	/* Threads to manage distribute transaction */
+	struct task_struct	*tdtd_commit_task;
+	atomic_t		tdtd_refcount;
+	struct lu_env		tdtd_env;
+
+	/* recovery update */
+	distribute_txn_replay_handler_t	tdtd_replay_handler;
+	struct list_head		tdtd_replay_list;
+	struct list_head		tdtd_replay_finish_list;
+	spinlock_t			tdtd_replay_list_lock;
+	/* last replay update transno */
+	__u32				tdtd_replay_ready:1;
+
+	/* Manage the llog recovery threads */
+	atomic_t		tdtd_recovery_threads_count;
+	wait_queue_head_t	tdtd_recovery_threads_waitq;
+	target_show_update_logs_retrievers_t
+				tdtd_show_update_logs_retrievers;
+	void			*tdtd_show_retrievers_cbdata;
+};
+
+struct tg_grants_data {
+	/* grants: all values in bytes */
+	/* grant lock to protect all grant counters */
+	spinlock_t		 tgd_grant_lock;
+	/* total amount of dirty data reported by clients in incoming obdo */
+	u64			 tgd_tot_dirty;
+	/* sum of filesystem space granted to clients for async writes */
+	u64			 tgd_tot_granted;
+	/* grant used by I/Os in progress (between prepare and commit) */
+	u64			 tgd_tot_pending;
+	/* amount of available space in percentage that is never used for
+	 * grants, used on MDT to always keep space for metadata. */
+	u64			 tgd_reserved_pcnt;
+	/* number of clients using grants */
+	int			 tgd_tot_granted_clients;
+	/* shall we grant space to clients not
+	 * supporting OBD_CONNECT_GRANT_PARAM? */
+	int			 tgd_grant_compat_disable;
+	/* protect all statfs-related counters */
+	spinlock_t		 tgd_osfs_lock;
+	time64_t		 tgd_osfs_age;
+	int			 tgd_blockbits;
+	/* counters used during statfs update, protected by ofd_osfs_lock.
+	 * record when some statfs refresh are in progress */
+	int			 tgd_statfs_inflight;
+	/* writes between prep & commit which might be accounted twice in
+	 * ofd_osfs.os_bavail */
+	u64			 tgd_osfs_unstable;
+	/* track writes completed while statfs refresh is underway.
+	 * tracking is only effective when ofd_statfs_inflight > 1 */
+	u64			 tgd_osfs_inflight;
+	/* statfs optimization: we cache a bit  */
+	struct obd_statfs	 tgd_osfs;
+};
+
 struct lu_target {
 	struct obd_device	*lut_obd;
 	struct dt_device	*lut_bottom;
+	struct dt_device_param	 lut_dt_conf;
+
+	struct target_distribute_txn_data *lut_tdtd;
 
 	/* supported opcodes and handlers for this target */
 	struct tgt_opc_slice	*lut_slice;
@@ -56,13 +168,12 @@ struct lu_target {
 	rwlock_t		 lut_sptlrpc_lock;
 	struct sptlrpc_rule_set	 lut_sptlrpc_rset;
 	spinlock_t		 lut_flags_lock;
-	int			 lut_sec_level;
-	unsigned int		 lut_mds_capa:1,
-				 lut_oss_capa:1,
-				 lut_syncjournal:1,
+	unsigned int		 lut_syncjournal:1,
 				 lut_sync_lock_cancel:2,
 				 /* e.g. OST node */
-				 lut_no_reconstruct:1;
+				 lut_no_reconstruct:1,
+				 /* enforce recovery for local clients */
+				 lut_local_recovery:1;
 	/** last_rcvd file */
 	struct dt_object	*lut_last_rcvd;
 	/* transaction callbacks */
@@ -77,6 +188,56 @@ struct lu_target {
 	spinlock_t		 lut_client_bitmap_lock;
 	/** Bitmap of known clients */
 	unsigned long		*lut_client_bitmap;
+	/* Number of clients supporting multiple modify RPCs
+	 * recorded in the bitmap */
+	atomic_t		 lut_num_clients;
+	/* Client generation to identify client slot reuse */
+	atomic_t		 lut_client_generation;
+	/** reply_data file */
+	struct dt_object	*lut_reply_data;
+	/** Bitmap of used slots in the reply data file */
+	unsigned long		**lut_reply_bitmap;
+	/** target sync count, used for debug & test */
+	atomic_t		 lut_sync_count;
+
+	/** cross MDT locks which should trigger Sync-on-Lock-Cancel */
+	spinlock_t		 lut_slc_locks_guard;
+	struct list_head	 lut_slc_locks;
+
+	/* target grants fields */
+	struct tg_grants_data	 lut_tgd;
+
+	/* target tunables */
+	const struct attribute	**lut_attrs;
+
+	/* FMD (file modification data) values */
+	int			 lut_fmd_max_num;
+	time64_t		 lut_fmd_max_age;
+};
+
+#define LUT_FMD_MAX_NUM_DEFAULT 128
+#define LUT_FMD_MAX_AGE_DEFAULT (obd_timeout + 10)
+
+/* number of slots in reply bitmap */
+#define LUT_REPLY_SLOTS_PER_CHUNK (1<<20)
+#define LUT_REPLY_SLOTS_MAX_CHUNKS 16
+
+#define TRD_INDEX_MEMORY -1
+
+/**
+ * Target reply data
+ */
+struct tg_reply_data {
+	/** chain of reply data anchored in tg_export_data */
+	struct list_head	trd_list;
+	/** copy of on-disk reply data */
+	struct lsd_reply_data	trd_reply;
+	/** versions for Version Based Recovery */
+	__u64			trd_pre_versions[4];
+	/** slot index in reply_data file */
+	int			trd_index;
+	/** tag the client used */
+	__u16			trd_tag;
 };
 
 extern struct lu_context_key tgt_session_key;
@@ -121,6 +282,11 @@ struct tgt_session_info {
 	bool			 tsi_preprocessed;
 	/* request JobID */
 	char                    *tsi_jobid;
+
+	/* update replay */
+	__u64			tsi_xid;
+	__u32			tsi_result;
+	__u32			tsi_client_gen;
 };
 
 static inline struct tgt_session_info *tgt_ses_info(const struct lu_env *env)
@@ -171,30 +337,22 @@ enum tgt_handler_flags {
 	/*
 	 * struct *_body is passed in the incoming message, and object
 	 * identified by this fid exists on disk.
-	 *                            *
-	 * "habeo corpus" == "I have a body"
 	 */
-	HABEO_CORPUS = (1 << 0),
+	HAS_BODY	= BIT(0),
 	/*
 	 * struct ldlm_request is passed in the incoming message.
-	 *
-	 * "habeo clavis" == "I have a key"
-	 *                                     */
-	HABEO_CLAVIS = (1 << 1),
+	 */
+	HAS_KEY		= BIT(1),
 	/*
 	 * this request has fixed reply format, so that reply message can be
 	 * packed by generic code.
-	 *
-	 * "habeo refero" == "I have a reply"
 	 */
-	HABEO_REFERO = (1 << 2),
+	HAS_REPLY	= BIT(2),
 	/*
 	 * this request will modify something, so check whether the file system
 	 * is readonly or not, then return -EROFS to client asap if necessary.
-	 *
-	 * "mutabor" == "I shall modify"
 	 */
-	MUTABOR      = (1 << 3)
+	IS_MUTABLE	= BIT(3)
 };
 
 struct tgt_handler {
@@ -207,7 +365,7 @@ struct tgt_handler {
 	/* Flags in enum tgt_handler_flags */
 	__u32			 th_flags;
 	/* Request version for this opcode */
-	int			 th_version;
+	enum lustre_msg_version	 th_version;
 	/* Handler function */
 	int			(*th_act)(struct tgt_session_info *tsi);
 	/* Handler function for high priority requests */
@@ -239,13 +397,23 @@ static inline int req_is_replay(struct ptlrpc_request *req)
 	return !!(lustre_msg_get_flags(req->rq_reqmsg) & MSG_REPLAY);
 }
 
+static inline bool tgt_is_multimodrpcs_client(struct obd_export *exp)
+{
+	return exp_connect_flags(exp) & OBD_CONNECT_MULTIMODRPCS;
+}
+
+static inline bool tgt_is_increasing_xid_client(struct obd_export *exp)
+{
+	return exp_connect_flags2(exp) & OBD_CONNECT2_INC_XID;
+}
+
 /* target/tgt_handler.c */
 int tgt_request_handle(struct ptlrpc_request *req);
 char *tgt_name(struct lu_target *tgt);
 void tgt_counter_incr(struct obd_export *exp, int opcode);
 int tgt_connect_check_sptlrpc(struct ptlrpc_request *req,
 			      struct obd_export *exp);
-int tgt_adapt_sptlrpc_conf(struct lu_target *tgt, int initial);
+int tgt_adapt_sptlrpc_conf(struct lu_target *tgt);
 int tgt_connect(struct tgt_session_info *tsi);
 int tgt_disconnect(struct tgt_session_info *uti);
 int tgt_obd_ping(struct tgt_session_info *tsi);
@@ -254,8 +422,6 @@ int tgt_convert(struct tgt_session_info *tsi);
 int tgt_bl_callback(struct tgt_session_info *tsi);
 int tgt_cp_callback(struct tgt_session_info *tsi);
 int tgt_llog_open(struct tgt_session_info *tsi);
-int tgt_llog_close(struct tgt_session_info *tsi);
-int tgt_llog_destroy(struct tgt_session_info *tsi);
 int tgt_llog_read_header(struct tgt_session_info *tsi);
 int tgt_llog_next_block(struct tgt_session_info *tsi);
 int tgt_llog_prev_block(struct tgt_session_info *tsi);
@@ -263,6 +429,7 @@ int tgt_sec_ctx_init(struct tgt_session_info *tsi);
 int tgt_sec_ctx_init_cont(struct tgt_session_info *tsi);
 int tgt_sec_ctx_fini(struct tgt_session_info *tsi);
 int tgt_sendpage(struct tgt_session_info *tsi, struct lu_rdpg *rdpg, int nob);
+int tgt_send_buffer(struct tgt_session_info *tsi, struct lu_rdbuf *rdbuf);
 int tgt_validate_obdo(struct tgt_session_info *tsi, struct obdo *oa);
 int tgt_sync(const struct lu_env *env, struct lu_target *tgt,
 	     struct dt_object *obj, __u64 start, __u64 end);
@@ -270,24 +437,29 @@ int tgt_sync(const struct lu_env *env, struct lu_target *tgt,
 int tgt_io_thread_init(struct ptlrpc_thread *thread);
 void tgt_io_thread_done(struct ptlrpc_thread *thread);
 
-int tgt_extent_lock(struct ldlm_namespace *ns, struct ldlm_res_id *res_id,
-		    __u64 start, __u64 end, struct lustre_handle *lh,
-		    int mode, __u64 *flags);
-void tgt_extent_unlock(struct lustre_handle *lh, ldlm_mode_t mode);
-int tgt_brw_lock(struct ldlm_namespace *ns, struct ldlm_res_id *res_id,
-		 struct obd_ioobj *obj, struct niobuf_remote *nb,
-		 struct lustre_handle *lh, int mode);
-void tgt_brw_unlock(struct obd_ioobj *obj, struct niobuf_remote *niob,
-		    struct lustre_handle *lh, int mode);
+int tgt_mdt_data_lock(struct ldlm_namespace *ns, struct ldlm_res_id *res_id,
+		      struct lustre_handle *lh, int mode, __u64 *flags);
+void tgt_mdt_data_unlock(struct lustre_handle *lh, enum ldlm_mode mode);
+int tgt_extent_lock(const struct lu_env *env, struct ldlm_namespace *ns,
+		    struct ldlm_res_id *res_id, __u64 start, __u64 end,
+		    struct lustre_handle *lh, int mode, __u64 *flags);
+void tgt_extent_unlock(struct lustre_handle *lh, enum ldlm_mode mode);
 int tgt_brw_read(struct tgt_session_info *tsi);
 int tgt_brw_write(struct tgt_session_info *tsi);
 int tgt_hpreq_handler(struct ptlrpc_request *req);
+void tgt_register_lfsck_in_notify_local(int (*notify)(const struct lu_env *,
+						      struct dt_device *,
+						      struct lfsck_req_local *,
+						      struct thandle *));
 void tgt_register_lfsck_in_notify(int (*notify)(const struct lu_env *,
 						struct dt_device *,
 						struct lfsck_request *));
 void tgt_register_lfsck_query(int (*query)(const struct lu_env *,
 					   struct dt_device *,
-					   struct lfsck_request *));
+					   struct lfsck_request *,
+					   struct lfsck_reply *,
+					   struct lfsck_query *));
+int req_can_reconstruct(struct ptlrpc_request *req, struct tg_reply_data *trd);
 
 extern struct tgt_handler tgt_sec_ctx_handlers[];
 extern struct tgt_handler tgt_lfsck_handlers[];
@@ -309,9 +481,9 @@ int tgt_hpreq_handler(struct ptlrpc_request *req);
 
 /* target/tgt_main.c */
 void tgt_boot_epoch_update(struct lu_target *lut);
-int tgt_last_commit_cb_add(struct thandle *th, struct lu_target *lut,
-			   struct obd_export *exp, __u64 transno);
-int tgt_new_client_cb_add(struct thandle *th, struct obd_export *exp);
+void tgt_save_slc_lock(struct lu_target *lut, struct ldlm_lock *lock,
+		       __u64 transno);
+void tgt_discard_slc_lock(struct lu_target *lut, struct ldlm_lock *lock);
 int tgt_init(const struct lu_env *env, struct lu_target *lut,
 	     struct obd_device *obd, struct dt_device *dt,
 	     struct tgt_opc_slice *slice,
@@ -322,18 +494,104 @@ void tgt_client_free(struct obd_export *exp);
 int tgt_client_del(const struct lu_env *env, struct obd_export *exp);
 int tgt_client_add(const struct lu_env *env, struct obd_export *exp, int);
 int tgt_client_new(const struct lu_env *env, struct obd_export *exp);
-int tgt_client_data_read(const struct lu_env *env, struct lu_target *tg,
-			 struct lsd_client_data *lcd, loff_t *off, int index);
-int tgt_client_data_write(const struct lu_env *env, struct lu_target *tg,
-			  struct lsd_client_data *lcd, loff_t *off, struct thandle *th);
-int tgt_server_data_read(const struct lu_env *env, struct lu_target *tg);
-int tgt_server_data_write(const struct lu_env *env, struct lu_target *tg,
-			  struct thandle *th);
 int tgt_server_data_update(const struct lu_env *env, struct lu_target *tg,
 			   int sync);
-int tgt_truncate_last_rcvd(const struct lu_env *env, struct lu_target *tg,
-			   loff_t off);
+int tgt_reply_data_init(const struct lu_env *env, struct lu_target *tgt);
+int tgt_lookup_reply(struct ptlrpc_request *req, struct tg_reply_data *trd);
+int tgt_mk_reply_data(const struct lu_env *env, struct lu_target *tgt,
+		      struct tg_export_data *ted, struct ptlrpc_request *req,
+		      __u64 opdata, struct thandle *th, bool write_update,
+		      __u64 transno);
+struct tg_reply_data *tgt_lookup_reply_by_xid(struct tg_export_data *ted,
+					       __u64 xid);
+int tgt_tunables_init(struct lu_target *lut);
+void tgt_tunables_fini(struct lu_target *lut);
+
+/* target/tgt_grant.c */
+static inline int exp_grant_param_supp(struct obd_export *exp)
+{
+	return !!(exp_connect_flags(exp) & OBD_CONNECT_GRANT_PARAM);
+}
 
+/* Blocksize used for client not supporting OBD_CONNECT_GRANT_PARAM.
+ * That's 4KB=2^12 which is the biggest block size known to work whatever
+ * the client's page size is. */
+#define COMPAT_BSIZE_SHIFT 12
+
+void tgt_grant_sanity_check(struct obd_device *obd, const char *func);
+void tgt_grant_connect(const struct lu_env *env, struct obd_export *exp,
+		       struct obd_connect_data *data, bool new_conn);
+void tgt_grant_discard(struct obd_export *exp);
+void tgt_grant_prepare_read(const struct lu_env *env, struct obd_export *exp,
+			    struct obdo *oa);
+void tgt_grant_prepare_write(const struct lu_env *env, struct obd_export *exp,
+			     struct obdo *oa, struct niobuf_remote *rnb,
+			     int niocount);
+void tgt_grant_commit(struct obd_export *exp, unsigned long grant_used, int rc);
+int tgt_grant_commit_cb_add(struct thandle *th, struct obd_export *exp,
+			    unsigned long grant);
+long tgt_grant_create(const struct lu_env *env, struct obd_export *exp,
+		      s64 *nr);
+int tgt_statfs_internal(const struct lu_env *env, struct lu_target *lut,
+			struct obd_statfs *osfs, time64_t max_age,
+			int *from_cache);
+ssize_t tot_dirty_show(struct kobject *kobj, struct attribute *attr,
+		       char *buf);
+ssize_t tot_granted_show(struct kobject *kobj, struct attribute *attr,
+			 char *buf);
+ssize_t tot_pending_show(struct kobject *kobj, struct attribute *attr,
+			 char *buf);
+ssize_t grant_compat_disable_show(struct kobject *kobj, struct attribute *attr,
+				  char *buf);
+ssize_t grant_compat_disable_store(struct kobject *kobj,
+				   struct attribute *attr,
+				   const char *buffer, size_t count);
+#if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 16, 53, 0)
+ssize_t sync_lock_cancel_show(struct kobject *kobj,
+			      struct attribute *attr, char *buf);
+ssize_t sync_lock_cancel_store(struct kobject *kobj, struct attribute *attr,
+			       const char *buffer, size_t count);
+#endif
+
+/* FMD */
+void tgt_fmd_update(struct obd_export *exp, const struct lu_fid *fid,
+		    __u64 xid);
+bool tgt_fmd_check(struct obd_export *exp, const struct lu_fid *fid,
+		   __u64 xid);
+#ifdef DO_FMD_DROP
+void tgt_fmd_drop(struct obd_export *exp, const struct lu_fid *fid);
+#else
+#define tgt_fmd_drop(exp, fid) do {} while (0)
+#endif
+
+/* target/update_trans.c */
+int distribute_txn_init(const struct lu_env *env,
+			struct lu_target *lut,
+			struct target_distribute_txn_data *tdtd,
+			__u32 index);
+void distribute_txn_fini(const struct lu_env *env,
+			 struct target_distribute_txn_data *tdtd);
+
+/* target/update_recovery.c */
+int insert_update_records_to_replay_list(struct target_distribute_txn_data *,
+					 struct llog_update_record *,
+					 struct llog_cookie *, __u32);
+void dtrq_list_dump(struct target_distribute_txn_data *tdtd,
+		    unsigned int mask);
+void dtrq_list_destroy(struct target_distribute_txn_data *tdtd);
+int distribute_txn_replay_handle(struct lu_env *env,
+			   struct target_distribute_txn_data *tdtd,
+			   struct distribute_txn_replay_req *dtrq);
+__u64 distribute_txn_get_next_transno(struct target_distribute_txn_data *tdtd);
+struct distribute_txn_replay_req *
+distribute_txn_get_next_req(struct target_distribute_txn_data *tdtd);
+void dtrq_destroy(struct distribute_txn_replay_req *dtrq);
+struct distribute_txn_replay_req_sub *
+dtrq_sub_lookup(struct distribute_txn_replay_req *dtrq, __u32 mdt_index);
+struct distribute_txn_replay_req *
+distribute_txn_lookup_finish_list(struct target_distribute_txn_data *tdtd,
+				  __u64 transno);
+bool is_req_replayed_by_update(struct ptlrpc_request *req);
 enum {
 	ESERIOUS = 0x0001000
 };
@@ -356,18 +614,6 @@ static inline int is_serious(int rc)
 	return (rc < 0 && -rc & ESERIOUS);
 }
 
-/**
- * Do not return server-side uid/gid to remote client
- */
-static inline void tgt_drop_id(struct obd_export *exp, struct obdo *oa)
-{
-	if (unlikely(exp_connect_rmtclient(exp))) {
-		oa->o_uid = -1;
-		oa->o_gid = -1;
-		oa->o_valid &= ~(OBD_MD_FLUID | OBD_MD_FLGID);
-	}
-}
-
 /*
  * Unified target generic handers macros and generic functions.
  */