/* * libretroshare/src/services: p3turtle.h * * Services for RetroShare. * * Copyright 2009 by Cyril Soler * * This library is free software; you can redistribute it and/or * modify it under the terms of the GNU Library General Public * License Version 2 as published by the Free Software Foundation. * * This library is distributed in the hope that it will be useful, * but WITHOUT ANY WARRANTY; without even the implied warranty of * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU * Library General Public License for more details. * * You should have received a copy of the GNU Library General Public * License along with this library; if not, write to the Free Software * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 * USA. * * Please report all bugs and problems to "csoler@users.sourceforge.net". * */ //====================================== General setup of the router ===================================// // // This class implements the turtle hopping router. It basically serves as // - a cache of turtle tunnels which are the communicating ways between distant peers. // - turtle tunnels are either end-point tunnels, or transitory points, in which case items are just // re-serialized and passed on along the tunnel. // - turtle tunnels are dug on request when calling diggTurtleTunnel(const std::string& hash) // this command lets a trace in each peer along the tunnel of where // packets come from and where they should go. Doing so, once a tunnel is // dug, packets are directly forwarded to the correct peer. // - an entry point for search request from the interface // - search results, as they come back, are forwarded upwards with some additional info: // - depth // depth of the file. setup to 1 for immediate friends and 2 for long distance friends. // - hash // hash of the file found // - name // name of the file found // - search request id. // // // - when downloading: // - for a given hash, a set of starting tunnels is maintained. Transitory // tunnels are also maintained for other hashes as requested by distant // peers. // //============================================= Operations =============================================// // // A download session works as follows: // Initiation: // 1 - the user searches for files (turtle search), and selects one and clicks download. // 2 - In parallel: // - the ft module gets a request, and searches for peers to provide this using its search modules. // - the turtle router is informed that a turtle download will happen with the given hash, so // it initiates tunnels for this hash. // In a loop: // 3 - the ft module asks the hash to the turtle searchModule, and sends file requests to the pqi // interface of this module. // 4 - the turtle pqi interface forwards these requests to the turtle router, which sends them to // the correct peers, selecting randomly among all the possible tunnels for this hash. // 5 - when a file data packet gets back, the turtle router forwards it back to the file transfer module. // //================================ connexion to the file transfer module ===============================// // // The turtle router should provide the ft module with the necessary interface for asking files, and // retreiving data: // - a search module that responds with a given fake peer id for hash request for which it has tunnels. // - a pqi interface to ask for file data // - p3turtle sends back file data packets to the file transfer module // //========================================== Tunnel usage rules ========================================// // // Tunnels should be used according to their capacity. This is an unsolved problem as for now. // //======================================= Tunnel maintenance rules =====================================// // // P3turtle should derive from pqihandler, just as p3disc, so that newly connected peers should trigger // asking for new tunnels, and disconnecting peers should produce a close tunnel packet. To simplify this, // I maintain a time stamp in tunnels, that is updated each time a file data packet travels in the tunnel. // Doing so, if a tunnel is not used for some time, it just disapears. Additional rules apply: // // - when a peer A connects: // - initiate new tunnels for all active file hashes (go through the list of hashes) by // asking to A, for the same hash and the same source. Only report tunnels for which the destination // endpoint is different, which should not happen in fact, because of bouncing gards. // // - when a peer A disconnects. // - do nothing. // // - when receive open tunnel from A // - check whether it's a bouncing request. If yes, give up. // - check hash against local files. // if > 0 // return tunnel ok item. No need to go forward, as sub tunnels are not useful. // else // forward request to peers, notting source and hashes. // // - when receive tunnel ok from A // - no need to check whether we already have this tunnel, as bouncing gards prevent this. // - leave a trace for the tunnel, and send (forward) backward. // // Ids management: // - tunnel ids should be identical for requests between 2 same peers for the same file hash. // - tunnel ids should be asymetric // - tunnel requests should never be identical, to allow searching multiple times for the same string. // So: // - when issuing an open tunnel order, // - a random request id is generated and used for packet routing // - a partial tunnel id is build, which is unique to the pair (source,file hash) // - when tunnel_ok is sent back, the tunnel id is completed so that it is unique to the // triplet (source, destination, file hash). // // For these needs, tunnels are represented by: // - their file hash. Each tunnel is only designed for transferring a single and same file. // - their local endpoints id. These are the ids of the peers in direction to the source and destination. // - the tunnel id, which is unique to the triple hash+global source+global destination. // - there is a difference between source and destination in tunnels. The source is the file asker, the // destination is the file provider. This helps sorting tunnels. // - a timestamp, used for cleaning unused tunnels. // // The turtle router has: // - a list of search requests and where to bounce them back. // - a list of tunnel digging requests and where to bounce them, back. // - a list of active file hashes, for which is should constantly maintain tunnels. // - a list of active tunnels, some being transitory, some being endpoints. // // Turtle router entries: // - a function for performing turtle search // - a function for handling tunnels for a given file hash. // // Questions: // - should tunnels be re-used ? nope. The only useful case would be when two peers are exchanging files, which happens quite rarely. // // TODO: // - derive most packets from RsTurtleTunnelPacketItem, that can be handled // as raw data through a tunnel. This will avoid replicating most of the // tunnel handling code. #ifndef MRK_PQI_TURTLE_H #define MRK_PQI_TURTLE_H #include #include #include "pqi/pqinetwork.h" #include "pqi/pqi.h" #include "pqi/pqimonitor.h" #include "services/p3service.h" #include "ft/ftsearch.h" #include "rsiface/rsturtle.h" #include "rsturtleitem.h" class ftServer ; class p3ConnectMgr; class ftDataMultiplex; class RsSerialiser; static const int TURTLE_MAX_SEARCH_DEPTH = 6 ; // This class is used to keep trace of requests (searches and tunnels). // class TurtleRequestInfo { public: TurtlePeerId origin ; // where the request came from. uint32_t time_stamp ; // last time the tunnel was actually used. Used for cleaning old tunnels. }; class TurtleTunnel { public: TurtlePeerId local_src ; // where packets come from. Direction to the source. TurtlePeerId local_dst ; // where packets should go. Direction to the destination. TurtleFileHash hash; // for starting and ending tunnels only. Null otherwise. TurtleVirtualPeerId vpid; // same, but contains the virtual peer id for this tunnel. uint32_t time_stamp ; // last time the tunnel was actually used. Used for cleaning old tunnels. }; // This class keeps trace of the activity for the file hashes the turtle router is asked to monitor. // class TurtleFileHashInfo { public: std::vector tunnels ; // list of active tunnel ids for this file hash TurtleRequestId last_request ; // last request for the tunnels of this hash TurtleFileName name ; time_t time_stamp ; uint64_t size ; }; // Subclassing: // // Class | Brings what | Usage // -----------+------------------+------------------------------------------------------ // p3Service | sendItem() | handle packet sending/receiving to/from friend peers. // pqiMonitor | configChanged() | handle who's connecting/disconnecting to dig new tunnels // RsTurtle | start/stop file()| brings interface for turtle service // ftSearch | search() | used to allow searching for monitored files. // p3Config | ConfigChanged() | used to load/save .cfg file for turtle variales. // -----------+------------------+------------------------------------------------------ // class p3turtle: public p3Service, public pqiMonitor, public RsTurtle,/* public ftSearch */ public p3Config { public: p3turtle(p3ConnectMgr *cm,ftServer *m); // Lauches a search request through the pipes, and immediately returns // the request id, which will be further used by the gui to store results // as they come back. // virtual TurtleSearchRequestId turtleSearch(const std::string& string_to_match) ; virtual TurtleSearchRequestId turtleSearch(const LinearizedExpression& expr) ; // Initiates tunnel handling for the given file hash. tunnels. Launches // an exception if an error occurs during the initialization process. The // turtle router itself does not initiate downloads, it only maintains // tunnels for the given hash. The download should be driven by the file // transfer module. Maybe this function can do the whole thing: // - initiate tunnel handling // - send the file request to the file transfer module // - populate the file transfer module with the adequate pqi interface and search module. // // This function should be called in addition to ftServer::FileRequest() so that the turtle router // automatically provide tunnels for the file to download. // virtual void monitorFileTunnels(const std::string& name,const std::string& file_hash,uint64_t size) ; /// This should be called when canceling a file download, so that the turtle router stops /// handling tunnels for this file. /// virtual void stopMonitoringFileTunnels(const std::string& file_hash) ; /// get info about tunnels virtual void getInfo(std::vector >&, std::vector >&, std::vector >&, std::vector >&) const ; /************* from pqiMonitor *******************/ /// Informs the turtle router that some peers are (dis)connected. This should initiate digging new tunnels, /// and closing other tunnels. /// virtual void statusChange(const std::list &plist); /************* from pqiMonitor *******************/ /// This function does many things: /// - It handles incoming and outgoing packets /// - it sorts search requests and forwards search results upward. /// - it cleans unused (tunnel+search) requests. /// - it maintains the pool of tunnels, for each request file hash. /// virtual int tick(); /************* from p3Config *******************/ virtual RsSerialiser *setupSerialiser() ; virtual std::list saveList(bool& cleanup) ; virtual bool loadList(std::list load) ; /************* Communication with ftserver *******************/ /// Does the turtle router manages tunnels to this peer ? (this is not a /// real id, but a fake one, that the turtle router is capable of connecting with a tunnel id). virtual bool isTurtlePeer(const std::string& peer_id) const ; /// Examines the peer id, finds the turtle tunnel in it, and respond yes if the tunnel is ok and operational. bool isOnline(const std::string& peer_id) const ; /// Returns a unique peer id, corresponding to the given tunnel. std::string getTurtlePeerId(TurtleTunnelId tid) const ; /// returns the list of virtual peers for all tunnels. void getVirtualPeersList(std::list& list) ; /// Send a data request into the correct tunnel for the given file hash void sendDataRequest(const std::string& peerId, const std::string& hash, uint64_t size, uint64_t offset, uint32_t chunksize) ; /// Send file data into the correct tunnel for the given file hash void sendFileData(const std::string& peerId, const std::string& hash, uint64_t size, uint64_t baseoffset, uint32_t chunksize, void *data) ; /// Send a request for the chunk map of this file to the given peer void sendChunkMapRequest(const std::string& peerId, const std::string& hash) ; /// Send a chunk map of this file to the given peer void sendChunkMap(const std::string& peerId, const std::string& hash,const CompressedChunkMap& cmap) ; private: //--------------------------- Admin/Helper functions -------------------------// /// Generates a cyphered combination of ownId() and file hash uint32_t generatePersonalFilePrint(const TurtleFileHash&,bool) ; /// Generates a random uint32_t number. uint32_t generateRandomRequestId() ; /// Auto cleaning of unused tunnels, search requests and tunnel requests. void autoWash() ; //------------------------------ Tunnel handling -----------------------------// /// initiates tunnels from here to any peers having the given file hash TurtleRequestId diggTunnel(const TurtleFileHash& hash) ; /// adds info related to a new virtual peer. void addDistantPeer(const TurtleFileHash&, TurtleTunnelId) ; //----------------------------- Routing functions ----------------------------// /// Handle tunnel digging for current file hashes void manageTunnels() ; /// closes a given tunnel. Should be called with mutex set. void locked_closeTunnel(TurtleTunnelId tid) ; /// Main routing function int handleIncoming(); /// Generic routing function for all tunnel packets that derive from RsTurtleGenericTunnelItem void routeGenericTunnelItem(RsTurtleGenericTunnelItem *item) ; /// specific routing functions for handling particular packets. void handleSearchRequest(RsTurtleSearchRequestItem *item); void handleSearchResult(RsTurtleSearchResultItem *item); void handleTunnelRequest(RsTurtleOpenTunnelItem *item); void handleTunnelResult(RsTurtleTunnelOkItem *item); void handleRecvFileRequest(RsTurtleFileRequestItem *item); void handleRecvFileData(RsTurtleFileDataItem *item); void handleRecvFileMapRequest(RsTurtleFileMapRequestItem*); void handleRecvFileMap(RsTurtleFileMapItem*); //------ Functions connecting the turtle router to other components.----------// /// Performs a search calling local cache and search structure. void performLocalSearch(const std::string& match_string,std::list& result) ; /// Returns a search result upwards (possibly to the gui) void returnSearchResult(RsTurtleSearchResultItem *item) ; /// Returns true if the file with given hash is hosted locally. bool performLocalHashSearch(const TurtleFileHash& hash,FileInfo& info) ; //--------------------------- Local variables --------------------------------// /* data */ p3ConnectMgr *mConnMgr; ftServer *_ft_server ; ftController *_ft_controller ; mutable RsMutex mTurtleMtx; /// keeps trace of who emmitted a given search request std::map _search_requests_origins ; /// keeps trace of who emmitted a tunnel request std::map _tunnel_requests_origins ; /// stores adequate tunnels for each file hash locally managed std::map _incoming_file_hashes ; /// stores file info for each file we provide. std::map _outgoing_file_hashes ; /// local tunnels, stored by ids (Either transiting or ending). std::map _local_tunnels ; /// Peers corresponding to each tunnel. std::map _virtual_peers ; /// Hashes marked to be deleted. std::vector _hashes_to_remove ; time_t _last_clean_time ; time_t _last_tunnel_management_time ; time_t _last_tunnel_campaign_time ; std::list _online_peers; /// used to force digging new tunnels bool _force_digg_new_tunnels ; #ifdef P3TURTLE_DEBUG // debug function void dumpState() ; #endif }; #endif