// Copyright 2015 The Kyua Authors. // All rights reserved. // // Redistribution and use in source and binary forms, with or without // modification, are permitted provided that the following conditions are // met: // // * Redistributions of source code must retain the above copyright // notice, this list of conditions and the following disclaimer. // * Redistributions in binary form must reproduce the above copyright // notice, this list of conditions and the following disclaimer in the // documentation and/or other materials provided with the distribution. // * Neither the name of Google Inc. nor the names of its contributors // may be used to endorse or promote products derived from this software // without specific prior written permission. // // THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS // "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT // LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR // A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT // OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, // SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT // LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, // DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY // THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT // (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE // OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. /// \file utils/process/executor.hpp /// Multiprogrammed process executor with isolation guarantees. /// /// This module provides a mechanism to invoke more than one process /// concurrently while at the same time ensuring that each process is run /// in a clean container and in a "safe" work directory that gets cleaned /// up automatically on termination. /// /// The intended workflow for using this module is the following: /// /// 1) Initialize the executor using setup(). Keep the returned object /// around through the lifetime of the next operations. Only one /// instance of the executor can be alive at once. /// 2) Spawn one or more processes with spawn(). On the caller side, keep /// track of any per-process data you may need using the returned /// exec_handle, which is unique among the set of active processes. /// 3) Call wait() or wait_any() to wait for completion of a process started /// in the previous step. Repeat as desired. /// 4) Use the returned exit_handle object by wait() or wait_any() to query /// the status of the terminated process and/or to access any of its /// data files. /// 5) Invoke cleanup() on the exit_handle to wipe any stale data. /// 6) Invoke cleanup() on the object returned by setup(). /// /// It is the responsibility of the caller to ensure that calls to /// spawn() and spawn_followup() are balanced with wait() and wait_any() calls. /// /// Processes executed in this manner have access to two different "unique" /// directories: the first is the "work directory", which is an empty directory /// that acts as the subprocess' work directory; the second is the "control /// directory", which is the location where the in-process code may place files /// that are not clobbered by activities in the work directory. #if !defined(UTILS_PROCESS_EXECUTOR_HPP) #define UTILS_PROCESS_EXECUTOR_HPP #include "utils/process/executor_fwd.hpp" #include #include #include "utils/datetime_fwd.hpp" #include "utils/fs/path_fwd.hpp" #include "utils/optional.hpp" #include "utils/passwd_fwd.hpp" #include "utils/process/child_fwd.hpp" #include "utils/process/status_fwd.hpp" namespace utils { namespace process { namespace executor { namespace detail { extern const char* stdout_name; extern const char* stderr_name; extern const char* work_subdir; /// Shared reference counter. typedef std::shared_ptr< std::size_t > refcnt_t; void setup_child(const utils::optional< utils::passwd::user >, const utils::fs::path&, const utils::fs::path&); } // namespace detail /// Maintenance data held while a subprocess is being executed. /// /// This data structure exists from the moment a subprocess is executed via /// executor::spawn() to when it is cleaned up with exit_handle::cleanup(). /// /// The caller NEED NOT maintain this object alive for the execution of the /// subprocess. However, the PID contained in here can be used to match /// exec_handle objects with corresponding exit_handle objects via their /// original_pid() method. /// /// Objects of this type can be copied around but their implementation is /// shared. The implication of this is that only the last copy of a given exit /// handle will execute the automatic cleanup() on destruction. class exec_handle { struct impl; /// Pointer to internal implementation. std::shared_ptr< impl > _pimpl; friend class executor_handle; exec_handle(std::shared_ptr< impl >); public: ~exec_handle(void); int pid(void) const; utils::fs::path control_directory(void) const; utils::fs::path work_directory(void) const; const utils::fs::path& stdout_file(void) const; const utils::fs::path& stderr_file(void) const; }; /// Container for the data of a process termination. /// /// This handle provides access to the details of the process that terminated /// and serves as the owner of the remaining on-disk files. The caller is /// expected to call cleanup() before destruction to remove the on-disk state. /// /// Objects of this type can be copied around but their implementation is /// shared. The implication of this is that only the last copy of a given exit /// handle will execute the automatic cleanup() on destruction. class exit_handle { struct impl; /// Pointer to internal implementation. std::shared_ptr< impl > _pimpl; friend class executor_handle; exit_handle(std::shared_ptr< impl >); detail::refcnt_t state_owners(void) const; public: ~exit_handle(void); void cleanup(void); int original_pid(void) const; const utils::optional< utils::process::status >& status(void) const; const utils::optional< utils::passwd::user >& unprivileged_user(void) const; const utils::datetime::timestamp& start_time() const; const utils::datetime::timestamp& end_time() const; utils::fs::path control_directory(void) const; utils::fs::path work_directory(void) const; const utils::fs::path& stdout_file(void) const; const utils::fs::path& stderr_file(void) const; }; /// Handler for the livelihood of the executor. /// /// Objects of this type can be copied around (because we do not have move /// semantics...) but their implementation is shared. Only one instance of the /// executor can exist at any point in time. class executor_handle { struct impl; /// Pointer to internal implementation. std::shared_ptr< impl > _pimpl; friend executor_handle setup(void); executor_handle(void) throw(); utils::fs::path spawn_pre(void); exec_handle spawn_post(const utils::fs::path&, const utils::fs::path&, const utils::fs::path&, const utils::datetime::delta&, const utils::optional< utils::passwd::user >, std::auto_ptr< utils::process::child >); void spawn_followup_pre(void); exec_handle spawn_followup_post(const exit_handle&, const utils::datetime::delta&, std::auto_ptr< utils::process::child >); public: ~executor_handle(void); const utils::fs::path& root_work_directory(void) const; void cleanup(void); template< class Hook > exec_handle spawn(Hook, const datetime::delta&, const utils::optional< utils::passwd::user >, const utils::optional< utils::fs::path > = utils::none, const utils::optional< utils::fs::path > = utils::none); template< class Hook > exec_handle spawn_followup(Hook, const exit_handle&, const datetime::delta&); exit_handle wait(const exec_handle); exit_handle wait_any(void); exit_handle reap(const pid_t); void check_interrupt(void) const; }; executor_handle setup(void); } // namespace executor } // namespace process } // namespace utils #endif // !defined(UTILS_PROCESS_EXECUTOR_HPP)