X-Git-Url: http://git.vrable.net/?a=blobdiff_plain;f=scandir.cc;h=e25352794c1a28c004d073150672170013ee2759;hb=15f7cef45e82e54a039e5486f08134c4f9a05471;hp=4b8f27fa6fcef0ea4ff85b6752363d16268a9656;hpb=39d47d5934a7401320a0c98095cad11e5c8abe52;p=cumulus.git diff --git a/scandir.cc b/scandir.cc index 4b8f27f..e253527 100644 --- a/scandir.cc +++ b/scandir.cc @@ -14,39 +14,100 @@ #include #include +#include "store.h" +#include "sha1.h" + using std::string; using std::vector; +static SegmentStore *segment_store; +static OutputStream *info_dump = 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 dumpfile(int fd) +/* Converts time to microseconds since the epoch. */ +int64_t encode_time(time_t time) +{ + return (int64_t)time * 1000000; +} + +/* 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) { struct stat stat_buf; fstat(fd, &stat_buf); int64_t size = 0; - char buf[4096]; - 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. */ + struct uuid segment_uuid; + int object_id; + OutputStream *index_data = index_segment->new_object(&segment_uuid, + &object_id, + "DREF"); + + 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) { + struct uuid block_segment_uuid; + int block_object_id; + + size_t bytes = file_read(fd, block_buf, LBS_BLOCK_SIZE); + if (bytes == 0) break; - } else { - size += res; - } + + hash.process(block_buf, bytes); + OutputStream *block = data_segment->new_object(&block_segment_uuid, + &block_object_id, + "DATA"); + block->write(block_buf, bytes); + index_data->write_uuid(block_segment_uuid); + index_data->write_u32(block_object_id); + + size += bytes; } - printf("Bytes read: %Ld\n", size); + file_info["sha1"] = string((const char *)hash.checksum(), + hash.checksum_size()); + file_info["data"] = encode_objref(segment_uuid, object_id); } void scanfile(const string& path) @@ -54,24 +115,64 @@ void scanfile(const string& path) int fd; long flags; struct stat stat_buf; + char *buf; + ssize_t len; + + // Set to true if the item is a directory and we should recursively scan + bool recurse = false; + + dictionary file_info; + lstat(path.c_str(), &stat_buf); - printf("%s:\n", path.c_str()); - printf(" ino=%Ld, perm=%04o, uid=%d, gid=%d, nlink=%d, blksize=%d, size=%Ld\n", - (int64_t)stat_buf.st_ino, stat_buf.st_mode & 07777, - stat_buf.st_uid, stat_buf.st_gid, stat_buf.st_nlink, - (int)stat_buf.st_blksize, (int64_t)stat_buf.st_size); + 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); + + char inode_type; switch (stat_buf.st_mode & S_IFMT) { case S_IFIFO: + inode_type = 'p'; + break; case S_IFSOCK: + inode_type = 's'; + break; case S_IFCHR: + inode_type = 'c'; + break; case S_IFBLK: + inode_type = 'b'; + break; case S_IFLNK: - printf(" special file\n"); + inode_type = 'l'; + + /* Use the reported file size to allocate a buffer large enough to read + * the symlink. Allocate slightly more space, so that we ask for more + * bytes than we expect and so check for truncation. */ + buf = new char[stat_buf.st_size + 2]; + len = readlink(path.c_str(), buf, stat_buf.st_size + 1); + if (len < 0) { + printf("error reading symlink: %m\n"); + } else if (len <= stat_buf.st_size) { + buf[len] = '\0'; + printf(" contents=%s\n", buf); + } else if (len > stat_buf.st_size) { + printf("error reading symlink: name truncated\n"); + } + + file_info["contents"] = buf; + + delete[] buf; break; case S_IFREG: - printf(" regular file\n"); + inode_type = '-'; + /* Be paranoid when opening the file. We have no guarantee that the * file was not replaced between the stat() call above and the open() * call below, so we might not even be opening a regular file. That @@ -90,21 +191,34 @@ void scanfile(const string& path) flags = fcntl(fd, F_GETFL); fcntl(fd, F_SETFL, flags & ~O_NONBLOCK); - dumpfile(fd); + file_info["size"] = encode_u64(stat_buf.st_size); + dumpfile(fd, file_info); close(fd); break; case S_IFDIR: - printf(" directory\n"); - scandir(path); + inode_type = 'd'; + recurse = true; break; + + default: + fprintf(stderr, "Unknown inode type: mode=%x\n", stat_buf.st_mode); + return; } + + file_info["type"] = string(1, inode_type); + + info_dump->write_string(path); + info_dump->write_dictionary(file_info); + + // If we hit a directory, now that we've written the directory itself, + // recursively scan the directory. + if (recurse) + scandir(path); } void scandir(const string& path) { - printf("Scan directory: %s\n", path.c_str()); - DIR *dir = opendir(path.c_str()); if (dir == NULL) { @@ -126,7 +240,6 @@ void scandir(const string& path) for (vector::iterator i = contents.begin(); i != contents.end(); ++i) { const string& filename = *i; - printf(" d_name = '%s'\n", filename.c_str()); scanfile(path + "/" + filename); } @@ -135,7 +248,27 @@ void scandir(const string& path) int main(int argc, char *argv[]) { - scandir("."); + block_buf = new char[LBS_BLOCK_SIZE]; + + segment_store = new SegmentStore("."); + SegmentWriter *sw = segment_store->new_segment(); + info_dump = sw->new_object(NULL, "ROOT"); + + index_segment = new SegmentPartitioner(segment_store); + data_segment = new SegmentPartitioner(segment_store); + + string uuid = SegmentWriter::format_uuid(sw->get_uuid()); + printf("Backup UUID: %s\n", uuid.c_str()); + + try { + scanfile("."); + } catch (IOException e) { + fprintf(stderr, "IOException: %s\n", e.getError().c_str()); + } + + delete index_segment; + delete data_segment; + delete sw; return 0; }