X-Git-Url: http://git.vrable.net/?a=blobdiff_plain;f=store.h;h=1f1edf8c590e7a3071ff3349545b6a50c4a6823e;hb=52df48ca169e07caa5c726d51ed4ed83aed748a5;hp=a9d6365c4525e271e498275f6a4aa93397a499f3;hpb=08147fb0bc1d09610b05f3f7d54eda88ef5776ec;p=cumulus.git diff --git a/store.h b/store.h index a9d6365..1f1edf8 100644 --- a/store.h +++ b/store.h @@ -1,16 +1,14 @@ /* LBS: An LFS-inspired filesystem backup system - * Copyright (C) 2006 Michael Vrable + * Copyright (C) 2006-2008 Michael Vrable * * Backup data is stored in a collection of objects, which are grouped together * into segments for storage purposes. This implementation of the object store - * is built on top of libtar, and represents segments as TAR files and objects - * as files within them. */ + * represents segments as TAR files and objects as files within them. */ #ifndef _LBS_STORE_H #define _LBS_STORE_H #include -#include #include #include @@ -19,7 +17,10 @@ #include #include +#include "localdb.h" +#include "remote.h" #include "sha1.h" +#include "ref.h" class LbsObject; @@ -40,54 +41,92 @@ public: std::string getError() const { return error; } }; +/* Simplified TAR header--we only need to store regular files, don't need to + * handle long filenames, etc. */ +static const int TAR_BLOCK_SIZE = 512; + +struct tar_header +{ + char name[100]; + char mode[8]; + char uid[8]; + char gid[8]; + char size[12]; + char mtime[12]; + char chksum[8]; + char typeflag; + char linkname[100]; + char magic[8]; + char uname[32]; + char gname[32]; + char devmajor[8]; + char devminor[8]; + char prefix[155]; + char padding[12]; +}; + /* A simple wrapper around a single TAR file to represent a segment. Objects * may only be written out all at once, since the tar header must be written * first; incremental writing is not supported. */ class Tarfile { public: - Tarfile(const std::string &path, const std::string &segment); + Tarfile(RemoteFile *file, const std::string &segment); ~Tarfile(); void write_object(int id, const char *data, size_t len); // Return an estimate of the size of the file. - size_t size_estimate() { return size; } - - void internal_write_object(const std::string &path, - const char *data, size_t len); + size_t size_estimate(); private: size_t size; std::string segment_name; - TAR *t; + + RemoteFile *file; + + /* Filter support. */ + int real_fd, filter_fd; + pid_t filter_pid; + + // Write data to the tar file + void tar_write(const char *data, size_t size); }; class TarSegmentStore { public: // New segments will be stored in the given directory. - TarSegmentStore(const std::string &path) { this->path = path; } + TarSegmentStore(RemoteStore *remote, + LocalDb *db = NULL) + { this->remote = remote; this->db = db; } ~TarSegmentStore() { sync(); } // Writes an object to segment in the store, and returns the name // (segment/object) to refer to it. The optional parameter group can be // used to control object placement; objects with different group // parameters are kept in separate segments. - std::string write_object(const char *data, size_t len, - const std::string &group = ""); + ObjectReference write_object(const char *data, size_t len, + const std::string &group = ""); // Ensure all segments have been fully written. void sync(); + // Dump statistics to stdout about how much data has been written + void dump_stats(); + private: struct segment_info { Tarfile *file; + std::string group; std::string name; // UUID - std::set refs; // Other segments this one refers to int count; // Objects written to this segment + int size; // Combined size of objects written + std::string basename; // Name of segment without directory + RemoteFile *rf; }; - std::string path; + RemoteStore *remote; std::map segments; + LocalDb *db; // Ensure that all segments in the given group have been fully written. void close_segment(const std::string &group); @@ -114,17 +153,21 @@ public: // incrementally. Data can be an arbitrary block of binary data of any // size. The pointer to the data need only remain valid until write() is // called. - //const char *get_data() const { return data; } - //size_t get_data_len() const { return data_len; } void set_data(const char *d, size_t len) { data = d; data_len = len; } // Write an object to a segment, thus making it permanent. This function // can be called at most once. void write(TarSegmentStore *store); + // Compute the checksum of an object, and include it in the object + // reference. This should be called after write(), and the data specified + // by set_data() must remain valid through the call to checksum(). + void checksum(); + // An object is assigned a permanent name once it has been written to a // segment. Until that time, its name cannot be determined. - std::string get_name() const { return name; } + std::string get_name() const { return ref.to_string(); } + ObjectReference get_ref() { return ref; } private: std::string group; @@ -132,9 +175,22 @@ private: size_t data_len; bool written; - std::string name; - - std::set refs; + ObjectReference ref; }; +/* Program through which segment data is piped before being written to file. */ +extern const char *filter_program; + +/* Extension which should be appended to segments written out (.tar is already + * included; this adds to it) */ +extern const char *filter_extension; + +/* Launch a process to filter data written to a file descriptor. fd_out is the + * file descriptor where the filtered data should be written. program is the + * filter program to execute (a single string which will be interpreted by + * /bin/sh). The return value is a file descriptor to which the data to be + * filtered should be written. The process ID of the filter process is stored + * at address filter_pid if non-NULL. */ +int spawn_filter(int fd_out, const char *program, pid_t *filter_pid); + #endif // _LBS_STORE_H