l_mdss_req_unlink_latency,
l_mdss_cap_revoke_eviction,
l_mdss_cap_acquisition_throttle,
+ l_mdss_req_getvxattr_latency,
l_mdss_last,
};
TRIM = (1<<2),
ENFORCE_LIVENESS = (1<<3),
};
+
explicit Server(MDSRank *m, MetricsHandler *metrics_handler);
~Server() {
g_ceph_context->get_perfcounters_collection()->remove(logger);
void _session_logged(Session *session, uint64_t state_seq, bool open, version_t pv,
const interval_set<inodeno_t>& inos_to_free, version_t piv,
const interval_set<inodeno_t>& inos_to_purge, LogSegment *ls);
- version_t prepare_force_open_sessions(map<client_t,entity_inst_t> &cm,
- map<client_t,client_metadata_t>& cmm,
- map<client_t,pair<Session*,uint64_t> >& smap);
- void finish_force_open_sessions(const map<client_t,pair<Session*,uint64_t> >& smap,
+ version_t prepare_force_open_sessions(std::map<client_t,entity_inst_t> &cm,
+ std::map<client_t,client_metadata_t>& cmm,
+ std::map<client_t,std::pair<Session*,uint64_t> >& smap);
+ void finish_force_open_sessions(const std::map<client_t,std::pair<Session*,uint64_t> >& smap,
bool dec_import=true);
- void flush_client_sessions(set<client_t>& client_set, MDSGatherBuilder& gather);
+ void flush_client_sessions(std::set<client_t>& client_set, MDSGatherBuilder& gather);
void finish_flush_session(Session *session, version_t seq);
void terminate_sessions();
void find_idle_sessions();
void kill_session(Session *session, Context *on_safe);
- size_t apply_blocklist(const std::set<entity_addr_t> &blocklist);
+ size_t apply_blocklist();
void journal_close_session(Session *session, int state, Context *on_safe);
size_t get_num_pending_reclaim() const { return client_reclaim_gather.size(); }
// -- requests --
void handle_client_request(const cref_t<MClientRequest> &m);
+ void handle_client_reply(const cref_t<MClientReply> &m);
void journal_and_reply(MDRequestRef& mdr, CInode *tracei, CDentry *tracedn,
LogEvent *le, MDSLogContextBase *fin);
// some helpers
bool check_fragment_space(MDRequestRef& mdr, CDir *in);
+ bool check_dir_max_entries(MDRequestRef& mdr, CDir *in);
bool check_access(MDRequestRef& mdr, CInode *in, unsigned mask);
bool _check_access(Session *session, CInode *in, unsigned mask, int caller_uid, int caller_gid, int setattr_uid, int setattr_gid);
CDentry *prepare_stray_dentry(MDRequestRef& mdr, CInode *in);
void journal_allocated_inos(MDRequestRef& mdr, EMetaBlob *blob);
void apply_allocated_inos(MDRequestRef& mdr, Session *session);
+ void _try_open_ino(MDRequestRef& mdr, int r, inodeno_t ino);
CInode* rdlock_path_pin_ref(MDRequestRef& mdr, bool want_auth,
bool no_want_auth=false);
CDentry* rdlock_path_xlock_dentry(MDRequestRef& mdr, bool create,
- bool okexist=false, bool want_layout=false);
+ bool okexist=false, bool authexist=false,
+ bool want_layout=false);
std::pair<CDentry*, CDentry*>
rdlock_two_paths_xlock_destdn(MDRequestRef& mdr, bool xlock_srcdn);
void handle_client_setlayout(MDRequestRef& mdr);
void handle_client_setdirlayout(MDRequestRef& mdr);
- int parse_quota_vxattr(string name, string value, quota_info_t *quota);
+ int parse_quota_vxattr(std::string name, std::string value, quota_info_t *quota);
void create_quota_realm(CInode *in);
- int parse_layout_vxattr(string name, string value, const OSDMap& osdmap,
+ int parse_layout_vxattr_json(std::string name, std::string value,
+ const OSDMap& osdmap, file_layout_t *layout);
+ int parse_layout_vxattr_string(std::string name, std::string value, const OSDMap& osdmap,
+ file_layout_t *layout);
+ int parse_layout_vxattr(std::string name, std::string value, const OSDMap& osdmap,
file_layout_t *layout, bool validate=true);
int check_layout_vxattr(MDRequestRef& mdr,
- string name,
- string value,
+ std::string name,
+ std::string value,
file_layout_t *layout);
void handle_set_vxattr(MDRequestRef& mdr, CInode *cur);
void handle_remove_vxattr(MDRequestRef& mdr, CInode *cur);
+ void handle_client_getvxattr(MDRequestRef& mdr);
void handle_client_setxattr(MDRequestRef& mdr);
void handle_client_removexattr(MDRequestRef& mdr);
void handle_client_fsync(MDRequestRef& mdr);
+ bool is_unlink_pending(CDentry *dn);
+ void wait_for_pending_unlink(CDentry *dn, MDRequestRef& mdr);
+
+ bool is_reintegrate_pending(CDentry *dn);
+ void wait_for_pending_reintegrate(CDentry *dn, MDRequestRef& mdr);
+
// open
void handle_client_open(MDRequestRef& mdr);
void handle_client_openc(MDRequestRef& mdr); // O_CREAT variant.
void handle_peer_link_prep_ack(MDRequestRef& mdr, const cref_t<MMDSPeerRequest> &m);
void do_link_rollback(bufferlist &rbl, mds_rank_t leader, MDRequestRef& mdr);
void _link_rollback_finish(MutationRef& mut, MDRequestRef& mdr,
- map<client_t,ref_t<MClientSnap>>& split);
+ std::map<client_t,ref_t<MClientSnap>>& split);
// unlink
void handle_client_unlink(MDRequestRef& mdr);
void _unlink_local_finish(MDRequestRef& mdr,
CDentry *dn, CDentry *straydn,
version_t);
- bool _rmdir_prepare_witness(MDRequestRef& mdr, mds_rank_t who, vector<CDentry*>& trace, CDentry *straydn);
+ bool _rmdir_prepare_witness(MDRequestRef& mdr, mds_rank_t who, std::vector<CDentry*>& trace, CDentry *straydn);
void handle_peer_rmdir_prep(MDRequestRef& mdr);
void _logged_peer_rmdir(MDRequestRef& mdr, CDentry *srcdn, CDentry *straydn);
void _commit_peer_rmdir(MDRequestRef& mdr, int r, CDentry *straydn);
void _renamesnap_finish(MDRequestRef& mdr, CInode *diri, snapid_t snapid);
// helpers
- bool _rename_prepare_witness(MDRequestRef& mdr, mds_rank_t who, set<mds_rank_t> &witnesse,
- vector<CDentry*>& srctrace, vector<CDentry*>& dsttrace, CDentry *straydn);
+ bool _rename_prepare_witness(MDRequestRef& mdr, mds_rank_t who, std::set<mds_rank_t> &witnesse,
+ std::vector<CDentry*>& srctrace, std::vector<CDentry*>& dsttrace, CDentry *straydn);
version_t _rename_prepare_import(MDRequestRef& mdr, CDentry *srcdn, bufferlist *client_map_bl);
bool _need_force_journal(CInode *diri, bool empty);
void _rename_prepare(MDRequestRef& mdr,
void _commit_peer_rename(MDRequestRef& mdr, int r, CDentry *srcdn, CDentry *destdn, CDentry *straydn);
void do_rename_rollback(bufferlist &rbl, mds_rank_t leader, MDRequestRef& mdr, bool finish_mdr=false);
void _rename_rollback_finish(MutationRef& mut, MDRequestRef& mdr, CDentry *srcdn, version_t srcdnpv,
- CDentry *destdn, CDentry *staydn, map<client_t,ref_t<MClientSnap>> splits[2],
+ CDentry *destdn, CDentry *staydn, std::map<client_t,ref_t<MClientSnap>> splits[2],
bool finish_mdr);
void evict_cap_revoke_non_responders();
bool terminating_sessions = false;
- set<client_t> client_reclaim_gather;
+ std::set<client_t> client_reclaim_gather;
+
+ const bufferlist& get_snap_trace(Session *session, SnapRealm *realm) const;
+ const bufferlist& get_snap_trace(client_t client, SnapRealm *realm) const;
private:
friend class MDSContinuation;
return xattr_name.rfind("ceph.dir.layout", 0) == 0 ||
xattr_name.rfind("ceph.file.layout", 0) == 0 ||
xattr_name.rfind("ceph.quota", 0) == 0 ||
- xattr_name == "ceph.dir.subvolume"sv ||
- xattr_name == "ceph.dir.pin"sv ||
- xattr_name == "ceph.dir.pin.random"sv ||
- xattr_name == "ceph.dir.pin.distributed"sv;
+ xattr_name == "ceph.dir.subvolume" ||
+ xattr_name == "ceph.dir.pin" ||
+ xattr_name == "ceph.dir.pin.random" ||
+ xattr_name == "ceph.dir.pin.distributed";
+ }
+
+ static bool is_ceph_dir_vxattr(std::string_view xattr_name) {
+ return (xattr_name == "ceph.dir.layout" ||
+ xattr_name == "ceph.dir.layout.json" ||
+ xattr_name == "ceph.dir.layout.object_size" ||
+ xattr_name == "ceph.dir.layout.stripe_unit" ||
+ xattr_name == "ceph.dir.layout.stripe_count" ||
+ xattr_name == "ceph.dir.layout.pool" ||
+ xattr_name == "ceph.dir.layout.pool_name" ||
+ xattr_name == "ceph.dir.layout.pool_id" ||
+ xattr_name == "ceph.dir.layout.pool_namespace" ||
+ xattr_name == "ceph.dir.pin" ||
+ xattr_name == "ceph.dir.pin.random" ||
+ xattr_name == "ceph.dir.pin.distributed");
+ }
+
+ static bool is_ceph_file_vxattr(std::string_view xattr_name) {
+ return (xattr_name == "ceph.file.layout" ||
+ xattr_name == "ceph.file.layout.json" ||
+ xattr_name == "ceph.file.layout.object_size" ||
+ xattr_name == "ceph.file.layout.stripe_unit" ||
+ xattr_name == "ceph.file.layout.stripe_count" ||
+ xattr_name == "ceph.file.layout.pool" ||
+ xattr_name == "ceph.file.layout.pool_name" ||
+ xattr_name == "ceph.file.layout.pool_id" ||
+ xattr_name == "ceph.file.layout.pool_namespace");
}
static bool is_allowed_ceph_xattr(std::string_view xattr_name) {
// before proceeding to reconnect_gather_finish
time reconnect_start = clock::zero();
time reconnect_last_seen = clock::zero();
- set<client_t> client_reconnect_gather; // clients i need a reconnect msg from.
- set<client_t> client_reconnect_denied; // clients whose reconnect msg have been denied .
+ std::set<client_t> client_reconnect_gather; // clients i need a reconnect msg from.
+ std::set<client_t> client_reconnect_denied; // clients whose reconnect msg have been denied .
feature_bitset_t supported_features;
+ feature_bitset_t supported_metric_spec;
feature_bitset_t required_client_features;
bool forward_all_requests_to_auth = false;
bool replay_unsafe_with_closed_session = false;
double cap_revoke_eviction_timeout = 0;
uint64_t max_snaps_per_dir = 100;
+ // long snapshot names have the following format: "_<SNAPSHOT-NAME>_<INODE-NUMBER>"
+ uint64_t snapshot_name_max = NAME_MAX - 1 - 1 - 13;
unsigned delegate_inos_pct = 0;
+ uint64_t dir_max_entries = 0;
+ int64_t bal_fragment_size_max = 0;
+
+ double inject_rename_corrupt_dentry_first = 0.0;
DecayCounter recall_throttle;
time last_recall_state;
double caps_throttle_retry_request_timeout;
size_t alternate_name_max = g_conf().get_val<Option::size_t>("mds_alternate_name_max");
+ size_t fscrypt_last_block_max_size = g_conf().get_val<Option::size_t>("mds_fscrypt_last_block_max_size");
};
static inline constexpr auto operator|(Server::RecallFlags a, Server::RecallFlags b) {