1 /* Cumulus: Efficient Filesystem Backup to the Cloud
2 * Copyright (C) 2006-2008 The Cumulus Developers
3 * See the AUTHORS file for a list of contributors.
5 * This program is free software; you can redistribute it and/or modify
6 * it under the terms of the GNU General Public License as published by
7 * the Free Software Foundation; either version 2 of the License, or
8 * (at your option) any later version.
10 * This program is distributed in the hope that it will be useful,
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13 * GNU General Public License for more details.
15 * You should have received a copy of the GNU General Public License along
16 * with this program; if not, write to the Free Software Foundation, Inc.,
17 * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
20 /* Backup data is stored in a collection of objects, which are grouped together
21 * into segments for storage purposes. This implementation of the object store
22 * represents segments as TAR files and objects as files within them. */
39 #include "third_party/sha1.h"
43 /* In memory datatype to represent key/value pairs of information, such as file
44 * metadata. Currently implemented as map<string, string>. */
45 typedef std::map<std::string, std::string> dictionary;
47 /* Simplified TAR header--we only need to store regular files, don't need to
48 * handle long filenames, etc. */
49 static const int TAR_BLOCK_SIZE = 512;
73 // It is valid for program to be NULL or empty; if so, no filtering is
75 static FileFilter *New(int fd, const char *program);
77 // Wait for the filter process to terminate.
80 // Accessors for the file descriptors.
81 int get_raw_fd() const { return fd_raw; }
82 int get_wrapped_fd() const { return fd_wrapped; }
85 FileFilter(int raw, int wrapped, pid_t pid);
87 // Launch a process to filter data written to a file descriptor. fd_out is
88 // the file descriptor where the filtered data should be written. program
89 // is the filter program to execute (a single string which will be
90 // interpreted by /bin/sh). The return value is a file descriptor to which
91 // the data to be filtered should be written. The process ID of the filter
92 // process is stored at address filter_pid if non-NULL.
93 static int spawn_filter(int fd_out, const char *program, pid_t *filter_pid);
95 // The original file descriptor passed when creating the FileFilter object.
98 // The wrapped file descriptor: writes here are piped through the filter
102 // The filter process if one was launched, or -1 if there is no filter
107 /* A simple wrapper around a single TAR file to represent a segment. Objects
108 * may only be written out all at once, since the tar header must be written
109 * first; incremental writing is not supported. */
112 Tarfile(RemoteFile *file, const std::string &segment);
115 void write_object(int id, const char *data, size_t len);
117 // Return an estimate of the size of the file.
118 size_t size_estimate();
122 std::string segment_name;
127 // Write data to the tar file
128 void tar_write(const char *data, size_t size);
131 class TarSegmentStore {
133 // New segments will be stored in the given directory.
134 TarSegmentStore(RemoteStore *remote,
136 { this->remote = remote; this->db = db; }
137 ~TarSegmentStore() { sync(); }
139 // Writes an object to segment in the store, and returns the name
140 // (segment/object) to refer to it. The optional parameter group can be
141 // used to control object placement; objects with different group
142 // parameters are kept in separate segments.
143 ObjectReference write_object(const char *data, size_t len,
144 const std::string &group = "",
145 const std::string &checksum = "",
148 // Ensure all segments have been fully written.
151 // Dump statistics to stdout about how much data has been written
155 struct segment_info {
158 std::string name; // UUID
159 int count; // Objects written to this segment
160 int data_size; // Combined size of objects written
161 std::string basename; // Name of segment without directory
166 std::map<std::string, struct segment_info *> segments;
169 // Ensure that all segments in the given group have been fully written.
170 void close_segment(const std::string &group);
172 // Parse an object reference string and return just the segment name
174 std::string object_reference_to_segment(const std::string &object);
177 /* An in-memory representation of an object, which can be incrementally built
178 * before it is written out to a segment. */
184 // If an object is placed in a group, it will be written out to segments
185 // only containing other objects in the same group. A group name is simply
187 //std::string get_group() const { return group; }
188 void set_group(const std::string &g) { group = g; }
190 // Data in an object must be written all at once, and cannot be generated
191 // incrementally. Data can be an arbitrary block of binary data of any
192 // size. The pointer to the data need only remain valid until write() is
193 // called. If checksum is non-NULL then it is assumed to contain a hash
194 // value for the data; this provides an optimization in case the caller has
195 // already checksummed the data. Otherwise the set_data will compute a
196 // hash of the data itself.
197 void set_data(const char *d, size_t len, const char *checksum);
199 // Explicitly sets the age of the data, for later garbage-collection or
200 // repacking purposes. If not set, the age defaults to the current time.
201 // The age is stored in the database as a floating point value, expressing
202 // the time in Julian days.
203 void set_age(double age) { this->age = age; }
205 // Write an object to a segment, thus making it permanent. This function
206 // can be called at most once.
207 void write(TarSegmentStore *store);
209 // An object is assigned a permanent name once it has been written to a
210 // segment. Until that time, its name cannot be determined.
211 ObjectReference get_ref() { return ref; }
218 std::string checksum;
224 /* Program through which segment data is piped before being written to file. */
225 extern const char *filter_program;
227 /* Extension which should be appended to segments written out (.tar is already
228 * included; this adds to it) */
229 extern const char *filter_extension;
231 #endif // _LBS_STORE_H