*/
/*
* This file is part of Lustre, http://www.lustre.org/
- * Lustre is a trademark of Sun Microsystems, Inc.
*
* lustre/osp/osp_internal.h
*
* those stale RPC(with older generation) will not be sent, otherwise it
* will cause update lllog corruption */
__u64 ou_generation;
-};
-struct osp_rpc_lock {
- /** Lock protecting in-flight RPC concurrency. */
- struct mutex rpcl_mutex;
- /** Used for MDS/RPC load testing purposes. */
- unsigned int rpcl_fakes;
+ /* dedicate update thread */
+ struct task_struct *ou_update_task;
+ struct lu_env ou_env;
};
struct osp_device {
* and required le64_to_cpu() conversion before use.
* Protected by opd_pre_lock */
struct lu_fid opd_last_used_fid;
+ /* on disk copy last_used_fid.f_oid or idif */
+ u64 opd_last_id;
struct lu_fid opd_gap_start_fid;
int opd_gap_count;
- /* connection to OST */
- struct osp_rpc_lock opd_rpc_lock;
struct obd_device *opd_obd;
struct obd_export *opd_exp;
- struct obd_uuid opd_cluuid;
struct obd_connect_data *opd_connect_data;
- int opd_connects;
+
/* connection status. */
unsigned int opd_new_connection:1,
opd_got_disconnected:1,
/* precreate structure for OSP */
struct osp_precreate *opd_pre;
/* dedicate precreate thread */
- struct ptlrpc_thread opd_pre_thread;
+ struct task_struct *opd_pre_task;
spinlock_t opd_pre_lock;
/* thread waits for signals about pool going empty */
wait_queue_head_t opd_pre_waitq;
/* send update thread */
struct osp_updates *opd_update;
- /* dedicate update thread */
- struct ptlrpc_thread opd_update_thread;
/*
* OST synchronization thread
struct llog_gen opd_sync_generation;
/* number of changes to sync, used to wake up sync thread */
atomic_t opd_sync_changes;
+ /* limit of changes to sync */
+ int opd_sync_max_changes;
/* processing of changes from previous mount is done? */
int opd_sync_prev_done;
/* found records */
- struct ptlrpc_thread opd_sync_thread;
+ struct task_struct *opd_sync_task;
wait_queue_head_t opd_sync_waitq;
/* list of in flight rpcs */
struct list_head opd_sync_in_flight_list;
*/
int opd_reserved_mb_high;
int opd_reserved_mb_low;
+ bool opd_cleanup_orphans_done;
+ bool opd_force_creation;
};
#define opd_pre_used_fid opd_pre->osp_pre_used_fid
* The left part is for value, binary mode. */
struct osp_xattr_entry {
struct list_head oxe_list;
- atomic_t oxe_ref;
void *oxe_value;
- size_t oxe_buflen;
- size_t oxe_namelen;
- size_t oxe_vallen;
- unsigned int oxe_exist:1,
- oxe_ready:1;
- char oxe_buf[0];
+ atomic_t oxe_ref;
+ unsigned int oxe_buflen;
+ unsigned int oxe_vallen;
+ unsigned short oxe_namelen;
+ unsigned short oxe_exist:1,
+ oxe_ready:1,
+ oxe_largebuf:1;
+ char oxe_name[0];
};
/* this is a top object */
struct dt_object opo_obj;
unsigned int opo_reserved:1,
opo_non_exist:1,
- opo_stale:1;
+ opo_stale:1,
+ opo_destroyed:1;
/* read/write lock for md osp object */
struct rw_semaphore opo_sem;
struct rw_semaphore opo_invalidate_sem;
};
-extern struct lu_object_operations osp_lu_obj_ops;
-extern const struct dt_device_operations osp_dt_ops;
-extern struct dt_object_operations osp_md_obj_ops;
-extern struct dt_body_operations osp_md_body_ops;
+extern const struct lu_object_operations osp_lu_obj_ops;
+extern const struct dt_object_operations osp_md_obj_ops;
+extern const struct dt_body_operations osp_md_body_ops;
struct osp_thread_info {
struct lu_buf osi_lb;
struct lu_attr osi_attr;
struct ost_id osi_oi;
struct ost_id osi_oi2;
- u64 osi_id;
loff_t osi_off;
union {
struct llog_rec_hdr osi_hdr;
static inline struct osp_thread_info *osp_env_info(const struct lu_env *env)
{
- struct osp_thread_info *info;
-
- info = lu_context_key_get(&env->le_ctx, &osp_thread_key);
- if (info == NULL) {
- lu_env_refill((struct lu_env *)env);
- info = lu_context_key_get(&env->le_ctx, &osp_thread_key);
- }
- LASSERT(info);
- return info;
+ return lu_env_info(env, &osp_thread_key);
}
struct osp_txn_info {
static inline struct osp_device *lu2osp_dev(struct lu_device *d)
{
LASSERT(lu_device_is_osp(d));
- return container_of0(d, struct osp_device, opd_dt_dev.dd_lu_dev);
+ return container_of_safe(d, struct osp_device, opd_dt_dev.dd_lu_dev);
}
static inline struct lu_device *osp2lu_dev(struct osp_device *d)
static inline struct osp_device *dt2osp_dev(struct dt_device *d)
{
LASSERT(lu_device_is_osp(&d->dd_lu_dev));
- return container_of0(d, struct osp_device, opd_dt_dev);
+ return container_of_safe(d, struct osp_device, opd_dt_dev);
}
static inline struct osp_object *lu2osp_obj(struct lu_object *o)
{
LASSERT(ergo(o != NULL, lu_device_is_osp(o->lo_dev)));
- return container_of0(o, struct osp_object, opo_obj.do_lu);
+ return container_of_safe(o, struct osp_object, opo_obj.do_lu);
}
static inline struct lu_object *osp2lu_obj(struct osp_object *obj)
static inline struct osp_object *osp_obj(const struct lu_object *o)
{
LASSERT(lu_device_is_osp(o->lo_dev));
- return container_of0(o, struct osp_object, opo_obj.do_lu);
+ return container_of_safe(o, struct osp_object, opo_obj.do_lu);
}
static inline struct osp_object *dt2osp_obj(const struct dt_object *d)
static inline struct dt_object *osp_object_child(struct osp_object *o)
{
- return container_of0(lu_object_next(osp2lu_obj(o)),
- struct dt_object, do_lu);
+ return container_of(lu_object_next(osp2lu_obj(o)),
+ struct dt_object, do_lu);
}
static inline struct seq_server_site *osp_seq_site(struct osp_device *osp)
return osp->opd_dt_dev.dd_lu_dev.ld_site->ld_seq_site;
}
-/**
- * Serializes in-flight MDT-modifying RPC requests to preserve idempotency.
- *
- * This mutex is used to implement execute-once semantics on the MDT.
- * The MDT stores the last transaction ID and result for every client in
- * its last_rcvd file. If the client doesn't get a reply, it can safely
- * resend the request and the MDT will reconstruct the reply being aware
- * that the request has already been executed. Without this lock,
- * execution status of concurrent in-flight requests would be
- * overwritten.
- *
- * This imlpementation limits the extent to which we can keep a full pipeline
- * of in-flight requests from a single client. This limitation can be
- * overcome by allowing multiple slots per client in the last_rcvd file,
- * see LU-6864.
- */
-#define OSP_FAKE_RPCL_IT ((void *)0x2c0012bfUL)
-
-static inline void osp_init_rpc_lock(struct osp_device *osp)
-{
- struct osp_rpc_lock *lck = &osp->opd_rpc_lock;
-
- mutex_init(&lck->rpcl_mutex);
- lck->rpcl_fakes = 0;
-}
-
-static inline void osp_get_rpc_lock(struct osp_device *osp)
-{
- struct osp_rpc_lock *lck = &osp->opd_rpc_lock;
-
- /* This would normally block until the existing request finishes.
- * If fail_loc is set it will block until the regular request is
- * done, then increment rpcl_fakes. Once that is non-zero it
- * will only be cleared when all fake requests are finished.
- * Only when all fake requests are finished can normal requests
- * be sent, to ensure they are recoverable again.
- */
- again:
- mutex_lock(&lck->rpcl_mutex);
-
- if (CFS_FAIL_CHECK_QUIET(OBD_FAIL_MDC_RPCS_SEM) ||
- CFS_FAIL_CHECK_QUIET(OBD_FAIL_OSP_RPCS_SEM)) {
- lck->rpcl_fakes++;
- mutex_unlock(&lck->rpcl_mutex);
-
- return;
- }
-
- /* This will only happen when the CFS_FAIL_CHECK() was just turned
- * off but there are still requests in progress. Wait until they
- * finish. It doesn't need to be efficient in this extremely rare
- * case, just have low overhead in the common case when it isn't true.
- */
- if (unlikely(lck->rpcl_fakes)) {
- mutex_unlock(&lck->rpcl_mutex);
- schedule_timeout(cfs_time_seconds(1) / 4);
-
- goto again;
- }
-}
-
-static inline void osp_put_rpc_lock(struct osp_device *osp)
-{
- struct osp_rpc_lock *lck = &osp->opd_rpc_lock;
-
- if (lck->rpcl_fakes) { /* OBD_FAIL_OSP_RPCS_SEM */
- mutex_lock(&lck->rpcl_mutex);
-
- if (lck->rpcl_fakes) /* check again under lock */
- lck->rpcl_fakes--;
- }
-
- mutex_unlock(&lck->rpcl_mutex);
-}
-
static inline int osp_fid_diff(const struct lu_fid *fid1,
const struct lu_fid *fid2)
{
fid_idif_id(fid2->f_seq, fid2->f_oid, 0);
}
- LASSERTF(fid_seq(fid1) == fid_seq(fid2), "fid1:"DFID
- ", fid2:"DFID"\n", PFID(fid1), PFID(fid2));
+ LASSERTF(fid_seq(fid1) == fid_seq(fid2), "fid1:"DFID", fid2:"DFID"\n",
+ PFID(fid1), PFID(fid2));
return fid_oid(fid1) - fid_oid(fid2);
}
+static inline void osp_fid_to_obdid(struct lu_fid *last_fid, u64 *osi_id)
+{
+ if (fid_is_idif((last_fid)))
+ *osi_id = fid_idif_id(fid_seq(last_fid), fid_oid(last_fid),
+ fid_ver(last_fid));
+ else
+ *osi_id = fid_oid(last_fid);
+}
static inline void osp_update_last_fid(struct osp_device *d, struct lu_fid *fid)
{
int diff = osp_fid_diff(fid, &d->opd_last_used_fid);
struct lu_fid *gap_start = &d->opd_gap_start_fid;
+
/*
* we might have lost precreated objects due to VBR and precreate
* orphans, the gap in objid can be calculated properly only here
PFID(&d->opd_gap_start_fid), d->opd_gap_count);
}
d->opd_last_used_fid = *fid;
+ osp_fid_to_obdid(fid, &d->opd_last_id);
}
}
ret; \
})
-static inline bool osp_send_update_thread_running(struct osp_device *osp)
-{
- return osp->opd_update_thread.t_flags & SVC_RUNNING;
-}
-
-static inline bool osp_send_update_thread_stopped(struct osp_device *osp)
-{
- return osp->opd_update_thread.t_flags & SVC_STOPPED;
-}
-
typedef int (*osp_update_interpreter_t)(const struct lu_env *env,
struct object_update_reply *rep,
struct ptlrpc_request *req,
int osp_xattr_del(const struct lu_env *env, struct dt_object *dt,
const char *name, struct thandle *th);
int osp_invalidate(const struct lu_env *env, struct dt_object *dt);
+bool osp_check_stale(struct dt_object *dt);
void osp_obj_invalidate_cache(struct osp_object *obj);
int osp_trans_stop(const struct lu_env *env, struct dt_device *dt,
/* osp_precreate.c */
int osp_init_precreate(struct osp_device *d);
-int osp_precreate_reserve(const struct lu_env *env, struct osp_device *d);
+int osp_precreate_reserve(const struct lu_env *env,
+ struct osp_device *d, bool can_block);
__u64 osp_precreate_get_id(struct osp_device *d);
int osp_precreate_get_fid(const struct lu_env *env, struct osp_device *d,
struct lu_fid *fid);
struct thandle *th);
/* lwp_dev.c */
-extern struct obd_ops lwp_obd_device_ops;
+extern const struct obd_ops lwp_obd_device_ops;
extern struct lu_device_type lwp_device_type;
static inline struct lu_device *osp2top(const struct osp_device *osp)