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;
71 /* A simple wrapper around a single TAR file to represent a segment. Objects
72 * may only be written out all at once, since the tar header must be written
73 * first; incremental writing is not supported. */
76 Tarfile(RemoteFile *file, const std::string &segment);
79 void write_object(int id, const char *data, size_t len);
81 // Return an estimate of the size of the file.
82 size_t size_estimate();
86 std::string segment_name;
91 int real_fd, filter_fd;
94 // Write data to the tar file
95 void tar_write(const char *data, size_t size);
98 class TarSegmentStore {
100 // New segments will be stored in the given directory.
101 TarSegmentStore(RemoteStore *remote,
103 { this->remote = remote; this->db = db; }
104 ~TarSegmentStore() { sync(); }
106 // Writes an object to segment in the store, and returns the name
107 // (segment/object) to refer to it. The optional parameter group can be
108 // used to control object placement; objects with different group
109 // parameters are kept in separate segments.
110 ObjectReference write_object(const char *data, size_t len,
111 const std::string &group = "",
112 const std::string &checksum = "",
115 // Ensure all segments have been fully written.
118 // Dump statistics to stdout about how much data has been written
122 struct segment_info {
125 std::string name; // UUID
126 int count; // Objects written to this segment
127 int data_size; // Combined size of objects written
128 std::string basename; // Name of segment without directory
133 std::map<std::string, struct segment_info *> segments;
136 // Ensure that all segments in the given group have been fully written.
137 void close_segment(const std::string &group);
139 // Parse an object reference string and return just the segment name
141 std::string object_reference_to_segment(const std::string &object);
144 /* An in-memory representation of an object, which can be incrementally built
145 * before it is written out to a segment. */
151 // If an object is placed in a group, it will be written out to segments
152 // only containing other objects in the same group. A group name is simply
154 //std::string get_group() const { return group; }
155 void set_group(const std::string &g) { group = g; }
157 // Data in an object must be written all at once, and cannot be generated
158 // incrementally. Data can be an arbitrary block of binary data of any
159 // size. The pointer to the data need only remain valid until write() is
160 // called. If checksum is non-NULL then it is assumed to contain a hash
161 // value for the data; this provides an optimization in case the caller has
162 // already checksummed the data. Otherwise the set_data will compute a
163 // hash of the data itself.
164 void set_data(const char *d, size_t len, const char *checksum);
166 // Explicitly sets the age of the data, for later garbage-collection or
167 // repacking purposes. If not set, the age defaults to the current time.
168 // The age is stored in the database as a floating point value, expressing
169 // the time in Julian days.
170 void set_age(double age) { this->age = age; }
172 // Write an object to a segment, thus making it permanent. This function
173 // can be called at most once.
174 void write(TarSegmentStore *store);
176 // An object is assigned a permanent name once it has been written to a
177 // segment. Until that time, its name cannot be determined.
178 ObjectReference get_ref() { return ref; }
185 std::string checksum;
191 /* Program through which segment data is piped before being written to file. */
192 extern const char *filter_program;
194 /* Extension which should be appended to segments written out (.tar is already
195 * included; this adds to it) */
196 extern const char *filter_extension;
198 /* Launch a process to filter data written to a file descriptor. fd_out is the
199 * file descriptor where the filtered data should be written. program is the
200 * filter program to execute (a single string which will be interpreted by
201 * /bin/sh). The return value is a file descriptor to which the data to be
202 * filtered should be written. The process ID of the filter process is stored
203 * at address filter_pid if non-NULL. */
204 int spawn_filter(int fd_out, const char *program, pid_t *filter_pid);
206 #endif // _LBS_STORE_H