From 3d8a08ca25c3041ac677335e51341d966a9b370b Mon Sep 17 00:00:00 2001 From: Amaury Pouly Date: Sun, 7 Feb 2016 21:42:15 +0000 Subject: hwstub: rewrite and expand library Rewrite the hwstub library in C++, with a clean and modular design. The library was designed from the ground up to be aware of multithreading issues and to handle memory allocation nicely with shared pointers. Compared to the original library, it brings the following major features: - support for JZ boot devices, it is very easy to add support for others - support for network transparent operations (through sockets): both tcp and unix domains are support Change-Id: I75899cb9c7aa938c17ede2bb3f468e7a55d625b4 --- utils/hwstub/include/hwstub_net.hpp | 334 ++++++++++++++++++++++++++++++++++++ 1 file changed, 334 insertions(+) create mode 100644 utils/hwstub/include/hwstub_net.hpp (limited to 'utils/hwstub/include/hwstub_net.hpp') diff --git a/utils/hwstub/include/hwstub_net.hpp b/utils/hwstub/include/hwstub_net.hpp new file mode 100644 index 0000000000..2cf6e07ccb --- /dev/null +++ b/utils/hwstub/include/hwstub_net.hpp @@ -0,0 +1,334 @@ +/*************************************************************************** + * __________ __ ___. + * Open \______ \ ____ ____ | | _\_ |__ _______ ___ + * Source | _// _ \_/ ___\| |/ /| __ \ / _ \ \/ / + * Jukebox | | ( <_> ) \___| < | \_\ ( <_> > < < + * Firmware |____|_ /\____/ \___ >__|_ \|___ /\____/__/\_ \ + * \/ \/ \/ \/ \/ + * $Id$ + * + * Copyright (C) 2016 by Amaury Pouly + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation; either version 2 + * of the License, or (at your option) any later version. + * + * This software is distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY + * KIND, either express or implied. + * + ****************************************************************************/ +#ifndef __HWSTUB_NET_HPP__ +#define __HWSTUB_NET_HPP__ + +#include "hwstub.hpp" +#include +#include +#include +#include + +namespace hwstub { +namespace net { + +/** Net context + * + * A socket context provides access to another context through a network. This + * is particularly useful to have another program create a USB context and provide + * access to it via some network. The two most useful types of network are TCP + * and Unix domains */ +class context : public hwstub::context +{ + friend class device; + friend class handle; +protected: + context(); +public: + virtual ~context(); + /** Create a socket context with an existing file descriptor. Note that the + * file descriptor will be closed when the context will be destroyed. */ + static std::shared_ptr create_socket(int socket_fd); + /** Create a TCP socket context with a domain name and a port. If port is empty, + * a default port is used. */ + static std::shared_ptr create_tcp(const std::string& domain, + const std::string& port, std::string *error = nullptr); + /** Create a UNIX socket context with a file system path (see man for details) */ + static std::shared_ptr create_unix(const std::string& path, + std::string *error = nullptr); + /** Create a UNIX socket context with an abstract name (see man for details) */ + static std::shared_ptr create_unix_abstract(const std::string& path, + std::string *error = nullptr); + /** Useful functions for network byte order conversion */ + uint32_t to_net_order(uint32_t u); + uint32_t from_net_order(uint32_t u); + + /** Default parameters */ + static std::string default_unix_path(); + static std::string default_tcp_domain(); + static std::string default_tcp_port(); + +protected: + /** Send a message to the server. Context will always serialize calls to send() + * so there is no need to worry about concurrency issues. */ + virtual error send(void *buffer, size_t& sz) = 0; + /** Receive a message from the server, sz is updated with the received size. + * Context will always serialize calls to recv() so there is no need to + * worry about concurrency issues. */ + virtual error recv(void *buffer, size_t& sz) = 0; + /** Perform a standard command: send a header with optional data and wait for + * an answer. In case of an underlying network error, the corresponding error + * code will be reported. If the server responds correctly, the argument array + * is overwritten with the servers's response. If the requests has been NACK'ed + * the error code will be parsed and returned as a standard error code (see details below) + * (note that the original error code can still be found in args[0]). No data + * is transmitted in case of NACK. + * If the server ACKs the request, this function will also perform reception of + * the data. In recv_data is not NULL, the receive data will be put there and the + * size will be written in in_size. There are two cases: either *recv_data is NULL + * and the function will allocate the memory based on much data is sent by the server. + * Or *recv_data is not NULL, in which case the function NOT allocate memory + * and put the data at *recv_data; in this case, *recv_size should be the set + * to the size of the buffer and will be updated to the received size. If the + * server sents more data than the buffer size, OVERFLOW will be returned. + * If no data was received but recv_data is not null, *recv_size will be set to + * zero. It is the caller's responsability to delete *recv_data. Note that if + * server sends data but recv_data is null, the data will still be received and + * thrown away. + * This function takes care of network byte order for cmd and arguments + * but not for data. */ + error send_cmd(uint32_t cmd, uint32_t args[HWSTUB_NET_ARGS], uint8_t *send_data, + size_t send_size, uint8_t **recv_data, size_t *recv_size); + /** Ask the context to stop any communication with the server and do a clean + * shutdown if possible. This is a blocking call. When this function returns, + * there will no more calls to the underlying communication functions. + * This function should be called in the destructor to prevent the context from + * calling children functions after the object has been deconstructed. */ + void stop_context(); + + /** Perform delayed init (aka HELLO stage), do nothing is not needed */ + void delayed_init(); + /* NOTE ctx_dev_t = uint32_t (device id) */ + uint32_t from_ctx_dev(ctx_dev_t dev); + ctx_dev_t to_ctx_dev(uint32_t dev); + virtual error fetch_device_list(std::vector& list, void*& ptr); + virtual void destroy_device_list(void *ptr); + virtual error create_device(ctx_dev_t dev, std::shared_ptr& hwdev); + virtual bool match_device(ctx_dev_t dev, std::shared_ptr hwdev); + + enum class state + { + HELLO, /* client is initialising, server has not been contacted yet */ + IDLE, /* not doing anything */ + DEAD, /* died on unrecoverable error */ + }; + + state m_state; /* client state */ + error m_error; /* error state for DEAD */ +}; + +/** Socket based net context + * + * Don't use this class directly, use context::create_* calls. This class + * provides send()/recv() for any socket based network. */ +class socket_context : public context +{ + friend class context; +protected: + socket_context(int socket_fd); +public: + virtual ~socket_context(); + /** set operation timeout */ + void set_timeout(std::chrono::milliseconds ms); + +protected: + virtual error send(void *buffer, size_t& sz); + virtual error recv(void *buffer, size_t& sz); + + int m_socketfd; /* socket file descriptor */ +}; + + +/** Net device + * + * Device accessed through a network */ +class device : public hwstub::device +{ + friend class context; /* for ctor */ +protected: + device(std::shared_ptr ctx, uint32_t devid); +public: + virtual ~device(); + +protected: + /** Return device ID */ + uint32_t device_id(); + virtual error open_dev(std::shared_ptr& handle); + virtual bool has_multiple_open() const; + + int32_t m_device_id; /* device id */ +}; + +/** Net handle + * + * Handle used to talk to a distant device. */ +class handle : public hwstub::handle +{ + friend class device; +protected: + handle(std::shared_ptr dev, uint32_t hid); +public: + virtual ~handle(); + +protected: + virtual error read_dev(uint32_t addr, void *buf, size_t& sz, bool atomic); + virtual error write_dev(uint32_t addr, const void *buf, size_t& sz, bool atomic); + virtual error get_dev_desc(uint16_t desc, void *buf, size_t& buf_sz); + virtual error get_dev_log(void *buf, size_t& buf_sz); + virtual error exec_dev(uint32_t addr, uint16_t flags); + virtual error status() const; + virtual size_t get_buffer_size(); + + uint32_t m_handle_id; /* handle id */ +}; + +/** Net server + * + * A server that forwards requests from net clients to a context */ +class server +{ +protected: + server(std::shared_ptr contex); +public: + virtual ~server(); + + /** Create a socket server with an existing file descriptor. Note that the + * file descriptor will be closed when the context will be destroyed. */ + static std::shared_ptr create_socket(std::shared_ptr contex, + int socket_fd); + /** Create a TCP socket server with a domain name and a port. If port is empty, + * a default port is used. */ + static std::shared_ptr create_tcp(std::shared_ptr contex, + const std::string& domain, const std::string& port, std::string *error = nullptr); + /** Create a UNIX socket server with a file system path (see man for details) */ + static std::shared_ptr create_unix(std::shared_ptr contex, + const std::string& path, std::string *error = nullptr); + /** Create a UNIX socket server with an abstract name (see man for details) */ + static std::shared_ptr create_unix_abstract( + std::shared_ptr contex, const std::string& path, + std::string *error = nullptr); + /** Useful functions for network byte order conversion */ + uint32_t to_net_order(uint32_t u); + uint32_t from_net_order(uint32_t u); + + /** Set/clear debug output for this context */ + void set_debug(std::ostream& os); + inline void clear_debug() { set_debug(cnull); } + /** Get debug output for this context */ + std::ostream& debug(); +protected: + struct client_state; + /** Opaque client type */ + typedef void* srv_client_t; + /** The client discovery implementation must call this function when a new + * client wants to talk to the server. If the server is unhappy with the + * request, it will immediately call terminate_client() */ + void client_arrived(srv_client_t client); + /** The client discovery implementation can notify asychronously about a client + * that left. Note that the implementation does not need to provide a mechanism, + * but should in this case return CLIENT_DISCONNECTED when the server performs + * a send() or recv() on a disconnected client. The server will always call + * after receiving client_left() but since this call is asychronous, the + * implementation must be prepared to deal with extra send()/recv() in the mean + * time. */ + void client_left(srv_client_t client); + /** The client discovery implementation can ask the server to stop all client + * threads. This is a blocking call. When this function returns, there will no + * more calls to the underlying communication functions. Note that the server + * will normally call terminate_client() on each active client at this point. + * This function should be called in the destructor to prevent the server from + * calling children functions after the object has been deconstructed. */ + void stop_server(); + /** Notify that the connection to a client is now finished. After this call, no + * more send()/recv() will be made to the client and the associated data will + * be freed. After this call, the implementation is not allowed to call client_left() + * for this client (assuming it did not previously). The implementation should close + * the communication channel at this point and free any associated data. */ + virtual void terminate_client(srv_client_t client) = 0; + /** Send a message to the client. Server will always serialize calls to send() + * for a given client so there is no need to worry about concurrency issues. */ + virtual error send(srv_client_t client, void *buffer, size_t& sz) = 0; + /** Receive a message from the client, sz is updated with the received size. + * Server will always serialize calls to recv() for a given client so there + * is no need to worry about concurrency issues. See comment about client_left(). */ + virtual error recv(srv_client_t client, void *buffer, size_t& sz) = 0; + /** handle command: cmd and arguments are in host order, the function should + * either return an error (command will be NACKed) or must fill the arguments + * and data for the answer. Note that the data is still in network byte order. + * If the funtion wants to send data back, it must set *send_data to a valid + * pointer, this pointer will be freed after the data is sent back. */ + error handle_cmd(client_state *state, uint32_t cmd, uint32_t args[HWSTUB_NET_ARGS], + uint8_t *recv_data, size_t recv_size, uint8_t*& send_data, size_t& send_size); + + /* complete state of a client */ + struct client_state + { + client_state(srv_client_t cl, std::future&& f); + srv_client_t client; /* client */ + std::future future; /* thread (see .cpp for explaination) */ + volatile bool exit; /* exit flag */ + uint32_t next_dev_id; /* next device ID */ + uint32_t next_handle_id; /* next handle ID */ + /* dev ID <-> hwstub dev map */ + std::map> dev_map; + /* handle ID -> hwstub handle map */ + std::map> handle_map; + }; + + /** Client thread */ + static void client_thread2(server *s, client_state *cs); + void client_thread(client_state *cs); + + std::shared_ptr m_context; /* context to perform operation */ + std::list m_client; /* client list */ + std::recursive_mutex m_mutex; /* server mutex */ + std::ostream *m_debug; /* debug stream */ +}; + +/** Socket based net server + * + */ +class socket_server : public server +{ +protected: + socket_server(std::shared_ptr contex, int socket_fd); +public: + virtual ~socket_server(); + /** create a server */ + static std::shared_ptr create(std::shared_ptr contex, + int socket_fd); + /** set operation timeout */ + void set_timeout(std::chrono::milliseconds ms); + +protected: + virtual void terminate_client(srv_client_t client); + virtual error send(srv_client_t client, void *buffer, size_t& sz); + virtual error recv(srv_client_t, void *buffer, size_t& sz); + + /* NOTE srv_client_t = int (client file descriptor) */ + int from_srv_client(srv_client_t cli); + srv_client_t to_srv_client(int fd); + + /** Discovery thread */ + static void discovery_thread1(socket_server *s); + void discovery_thread(); + + static const int LISTEN_QUEUE_SIZE = 5; + struct timeval m_timeout; /* operations timeout */ + int m_socketfd; /* socket file descriptor */ + std::thread m_discovery_thread; /* thread handling client discovery */ + volatile bool m_discovery_exit; /* exit flag */ +}; + +} // namespace net +} // namespace hwstub + +#endif /* __HWSTUB_NET_HPP__ */ -- cgit v1.2.3