X-Git-Url: http://git.vrable.net/?a=blobdiff_plain;f=scandir.cc;h=f3bf3027601c6932e981701499232d772d47392a;hb=d5afe2c08724b4b263a76f187e39861c1e6e0cf5;hp=043d35eeba637f2cba740c7ae0ef0d5ac5d96596;hpb=358b7ac6f741088838fb84979c9127976eb34fdf;p=cumulus.git diff --git a/scandir.cc b/scandir.cc index 043d35e..f3bf302 100644 --- a/scandir.cc +++ b/scandir.cc @@ -12,20 +12,27 @@ #include #include +#include #include +#include +#include +#include "format.h" #include "store.h" #include "sha1.h" +using std::list; using std::string; using std::vector; +using std::ostream; -static SegmentStore *segment_store; -static OutputStream *info_dump = NULL; +static TarSegmentStore *tss = NULL; -static SegmentPartitioner *index_segment, *data_segment; +/* Buffer for holding a single block of data read from a file. */ +static const int LBS_BLOCK_SIZE = 1024 * 1024; +static char *block_buf; -void scandir(const string& path); +void scandir(const string& path, std::ostream& metadata); /* Converts time to microseconds since the epoch. */ int64_t encode_time(time_t time) @@ -33,48 +40,98 @@ int64_t encode_time(time_t time) return (int64_t)time * 1000000; } -void dumpfile(int fd, dictionary &file_info) +/* Read data from a file descriptor and return the amount of data read. A + * short read (less than the requested size) will only occur if end-of-file is + * hit. */ +size_t file_read(int fd, char *buf, size_t maxlen) +{ + size_t bytes_read = 0; + + while (true) { + ssize_t res = read(fd, buf, maxlen); + if (res < 0) { + if (errno == EINTR) + continue; + throw IOException("file_read: error reading"); + } else if (res == 0) { + break; + } else { + bytes_read += res; + buf += res; + maxlen -= res; + } + } + + return bytes_read; +} + +/* Read the contents of a file (specified by an open file descriptor) and copy + * the data to the store. */ +void dumpfile(int fd, dictionary &file_info, ostream &metadata) { struct stat stat_buf; fstat(fd, &stat_buf); int64_t size = 0; - - char buf[4096]; + list segment_list; if ((stat_buf.st_mode & S_IFMT) != S_IFREG) { printf("file is no longer a regular file!\n"); return; } + /* The index data consists of a sequence of pointers to the data blocks + * that actually comprise the file data. This level of indirection is used + * so that the same data block can be used in multiple files, or multiple + * versions of the same file. */ SHA1Checksum hash; while (true) { - ssize_t res = read(fd, buf, sizeof(buf)); - if (res < 0) { - if (errno == EINTR) - continue; - printf("Error while reading: %m\n"); - return; - } else if (res == 0) { + size_t bytes = file_read(fd, block_buf, LBS_BLOCK_SIZE); + if (bytes == 0) break; - } else { - hash.process(buf, res); - OutputStream *block = data_segment->new_object(); - block->write(buf, res); - size += res; - } + + hash.process(block_buf, bytes); + + // tarstore processing + string blockid = tss->write_object(block_buf, bytes, "data"); + segment_list.push_back(blockid); + + size += bytes; } - file_info["sha1"] = string((const char *)hash.checksum(), - hash.checksum_size()); + file_info["checksum"] = hash.checksum_str(); + + /* For files that only need to be broken apart into a few objects, store + * the list of objects directly. For larger files, store the data + * out-of-line and provide a pointer to the indrect object. */ + if (segment_list.size() < 8) { + string blocklist = ""; + for (list::iterator i = segment_list.begin(); + i != segment_list.end(); ++i) { + if (i != segment_list.begin()) + blocklist += " "; + blocklist += *i; + } + file_info["data"] = blocklist; + } else { + string blocklist = ""; + for (list::iterator i = segment_list.begin(); + i != segment_list.end(); ++i) { + blocklist += *i + "\n"; + } + string indirect = tss->write_object(blocklist.data(), blocklist.size(), + "indirect"); + file_info["data"] = "@" + indirect; + } } -void scanfile(const string& path) +void scanfile(const string& path, ostream &metadata) { int fd; long flags; struct stat stat_buf; char *buf; ssize_t len; + list refs; // Set to true if the item is a directory and we should recursively scan bool recurse = false; @@ -85,12 +142,14 @@ void scanfile(const string& path) printf("%s\n", path.c_str()); - file_info["mode"] = encode_u16(stat_buf.st_mode & 07777); - file_info["atime"] = encode_u64(encode_time(stat_buf.st_atime)); - file_info["ctime"] = encode_u64(encode_time(stat_buf.st_ctime)); - file_info["mtime"] = encode_u64(encode_time(stat_buf.st_mtime)); - file_info["user"] = encode_u32(stat_buf.st_uid); - file_info["group"] = encode_u32(stat_buf.st_gid); + metadata << "name: " << uri_encode(path) << "\n"; + + file_info["mode"] = encode_int(stat_buf.st_mode & 07777); + file_info["atime"] = encode_int(stat_buf.st_atime); + file_info["ctime"] = encode_int(stat_buf.st_ctime); + file_info["mtime"] = encode_int(stat_buf.st_mtime); + file_info["user"] = encode_int(stat_buf.st_uid); + file_info["group"] = encode_int(stat_buf.st_gid); char inode_type; @@ -124,7 +183,7 @@ void scanfile(const string& path) printf("error reading symlink: name truncated\n"); } - file_info["contents"] = buf; + file_info["contents"] = uri_encode(buf); delete[] buf; break; @@ -149,8 +208,8 @@ void scanfile(const string& path) flags = fcntl(fd, F_GETFL); fcntl(fd, F_SETFL, flags & ~O_NONBLOCK); - file_info["size"] = encode_u64(stat_buf.st_size); - dumpfile(fd, file_info); + file_info["size"] = encode_int(stat_buf.st_size); + dumpfile(fd, file_info, metadata); close(fd); break; @@ -165,17 +224,18 @@ void scanfile(const string& path) } file_info["type"] = string(1, inode_type); + metadata << "type: " << inode_type << "\n"; - info_dump->write_string(path); - info_dump->write_dictionary(file_info); + dict_output(metadata, file_info); + metadata << "\n"; // If we hit a directory, now that we've written the directory itself, // recursively scan the directory. if (recurse) - scandir(path); + scandir(path, metadata); } -void scandir(const string& path) +void scandir(const string& path, ostream &metadata) { DIR *dir = opendir(path.c_str()); @@ -198,7 +258,7 @@ void scandir(const string& path) for (vector::iterator i = contents.begin(); i != contents.end(); ++i) { const string& filename = *i; - scanfile(path + "/" + filename); + scanfile(path + "/" + filename, metadata); } closedir(dir); @@ -206,25 +266,23 @@ void scandir(const string& path) int main(int argc, char *argv[]) { - segment_store = new SegmentStore("."); - SegmentWriter *sw = segment_store->new_segment(); - info_dump = sw->new_object(); + block_buf = new char[LBS_BLOCK_SIZE]; - index_segment = new SegmentPartitioner(segment_store); - data_segment = new SegmentPartitioner(segment_store); + tss = new TarSegmentStore("."); - string uuid = SegmentWriter::format_uuid(sw->get_uuid()); - printf("Backup UUID: %s\n", uuid.c_str()); + std::ostringstream metadata; try { - scanfile("."); + scanfile(".", metadata); } catch (IOException e) { fprintf(stderr, "IOException: %s\n", e.getError().c_str()); } - delete index_segment; - delete data_segment; - delete sw; + const string md = metadata.str(); + string root = tss->write_object(md.data(), md.size(), "root"); + + tss->sync(); + delete tss; return 0; }